An unmanned aerial vehicle hoisting and transporting swing suppression method, device, equipment and medium

Abstract

This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, and discloses a method, device, equipment, and medium for suppressing sway during UAV hoisting and transportation. The method includes acquiring a flight load image corresponding to the target load of the UAV during flight; acquiring a reference load image corresponding to the target load of the UAV in a stable hovering state; calculating a target control compensation vector corresponding to the target load of the UAV during flight based on the flight load image and the reference load image; and correcting the initial trajectory of the UAV based on the target control compensation vector to obtain the target trajectory of the UAV. This method generates a compensation vector to correct the flight trajectory through calculations on two types of load images, requires no additional sensors, can adapt to changing hoisting conditions, suppresses load sway in real time, reduces hardware costs, and improves flight stability and safety.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, specifically to a method, device, equipment, and medium for suppressing swaying during UAV hoisting and transportation. Background Technology

[0002] Current drone hoisting operations generally employ dynamic modeling control schemes, relying on external sensors such as inertial and laser rangefinders to obtain rope length and load swing status. This hardware assembly increases the drone's weight, limiting its endurance. Furthermore, this method is highly dependent on fixed physical parameters. In field operations, factors such as airflow interference, rope tension, and load changes cause continuous dynamic changes in rope length, leading to inaccurate model parameters, significant lag in disturbance estimation, and large deviations in controller output compensation. During high-speed flight and in strong winds, the load is prone to large swings, potentially even causing the aircraft to oscillate and lose control. Simultaneously, existing technologies lack solutions for real-time swing compensation calculation through load image comparison, making it impossible to achieve non-contact load status perception and autonomous trajectory correction solely through airborne vision. This results in poor trajectory tracking accuracy and operational safety in complex scenarios. Summary of the Invention

[0003] This invention provides a method, device, equipment, and medium for suppressing swing during unmanned aerial vehicle (UAV) hoisting and transportation, in order to solve the problems of existing UAV hoisting control relying on additional sensing equipment, difficulty in accurately identifying dynamic rope length and external disturbances, poor load swing suppression effect under complex field conditions, low trajectory tracking accuracy, and easy oscillation and loss of control during flight.

[0004] In a first aspect, the present invention provides a method for suppressing swaying during unmanned aerial vehicle (UAV) hoisting and transportation, the method comprising: Acquire flight payload images corresponding to the target payload of the target UAV during its flight; Acquire a reference load image corresponding to the target payload of the target UAV in a stable hovering state; Based on the flight load image and the reference load image, calculate the target control compensation vector corresponding to the target load of the target UAV during flight; Based on the target control compensation vector, the initial trajectory corresponding to the target UAV is corrected to obtain the target trajectory corresponding to the target UAV.

[0005] This method generates compensation vectors to correct flight trajectories through calculations on two types of load images. It requires no additional sensors, can adapt to varying hoisting conditions, suppress load sway in real time, reduce hardware costs, and improve flight stability and safety.

[0006] In one optional implementation, based on the flight load image and the reference load image, the target control compensation vector corresponding to the target load of the target UAV during flight is calculated, including: Target recognition is performed on the flight payload image to determine the target payload region and the payload outline pixel area in the flight payload image. Based on the target load region, calculate the real-time load swing angle corresponding to the target load of the target UAV during flight; Target recognition is performed on the reference load image to determine the reference load pixel area and reference physical rope length corresponding to the reference load image; Based on the payload outline pixel area, the reference payload pixel area, and the reference physical rope length, the dynamic rope length corresponding to the target payload of the target UAV during flight is calculated. The target control compensation vector is calculated based on the real-time swing angle of the load and the dynamic rope length.

[0007] This method achieves load visual recognition by comparing flight load images with reference load images. It can simultaneously calculate the real-time swing angle and dynamic rope length of the load using only the pixel information of the airborne images, without the need for additional hardware such as laser rangefinders and external inertial sensors, thus reducing the payload and equipment cost of the UAV. The method uses standard parameters calibrated from the reference image to correct the calculated rope length value in real time, which can adapt to the dynamic changes in rope length caused by working conditions such as rope deformation and variable load. The swing angle and rope length perception accuracy is higher. Based on this, the control compensation vector can accurately match the actual hoisting disturbance, which can significantly improve the load swing suppression effect and the accuracy of trajectory correction.

[0008] In one optional implementation, based on the target load region, the real-time load swing angle corresponding to the target load of the target UAV during flight is calculated, including: Obtain the center coordinates of the payload image corresponding to the flight payload image; Based on the target load region, determine the load pixel position coordinates corresponding to the geometric center of the target load; Based on the load pixel position coordinates and the load image center coordinates, a two-dimensional pixel deviation vector is calculated; Based on the lateral deviation vector in the two-dimensional pixel deviation vector, the real-time lateral swing angle corresponding to the target load of the target UAV during flight is calculated. Based on the longitudinal deviation vector in the two-dimensional pixel deviation vector, the real-time longitudinal sway angle corresponding to the target payload of the target UAV during flight is calculated. The real-time swing angle of the load is calculated based on the real-time lateral and longitudinal swing angles.

[0009] This method constructs a two-dimensional deviation vector by offsetting the pixel offset between the image center and the geometric center of the load. It then calculates the horizontal and vertical swing angles separately and fuses them to obtain the complete load swing angle. The two-dimensional swing state can be obtained non-contactly and synchronously using only visual pixel information, without the need for additional angle sensing devices. The pixel offset directly maps to the spatial swing angle. The calculation logic is simple and has strong real-time performance. It can accurately capture the swing offset of the load in any direction, providing reliable angle data support for the subsequent generation of high-precision control compensation vectors and improving the overall anti-swing control accuracy.

[0010] In one optional implementation, the target control compensation vector is calculated based on the real-time swing angle of the load and the dynamic rope length, including: Calculate the load swing angular velocity corresponding to the target load based on the real-time swing angle of the load; An augmented state vector is generated based on the real-time swing angle of the load, the dynamic rope length, and the angular velocity of the load swing. Configure the target gain matrix of the extended state observer based on the dimension of the augmented state vector; The augmented state vector is input into the extended state observer, and a discrete-time recursion is performed based on the target gain matrix to output the total perturbation estimate. The target control compensation vector is calculated based on the total disturbance estimate.

[0011] This method obtains the swing angular velocity by differentiating the swing angle, integrates the dynamic rope length, swing angle, and angular velocity to construct an augmented state vector, and then relies on the gain matrix of the matching dimension to drive the extended state observer for discrete recursion. It can uniformly identify various lumped disturbances such as rope length changes, load swing, and external wind disturbances. It can obtain a high-precision total disturbance estimate without the need for an accurate and complex UAV hoisting dynamics model. The target control compensation vector calculated in this way can accurately offset the flight interference caused by load swing, greatly improve the hoisting system's anti-disturbance capability and swing suppression effect, and has stronger adaptability to complex working conditions such as variable rope length and strong wind.

[0012] In one optional implementation, the target control compensation vector is calculated based on the total disturbance estimate, including: The total disturbance estimate is decomposed to obtain the horizontal equivalent disturbance component and the vertical equivalent disturbance component; Calculate the geometric mapping coefficients based on the dynamic rope length and gravitational acceleration; The geometric mapping coefficients are multiplied by the transverse equivalent disturbance component and the longitudinal equivalent disturbance component, respectively, to obtain the transverse original disturbance acceleration and the longitudinal original disturbance acceleration. The signs of the original lateral and longitudinal disturbance accelerations are inverted to obtain the corrected lateral and longitudinal disturbance accelerations. The acceleration correction vector is calculated based on the lateral and longitudinal corrected perturbation accelerations. Based on the acceleration correction vector, the target control compensation vector is obtained.

[0013] This method decomposes the total disturbance into independent horizontal and vertical components, and uses dynamic rope length and gravity to construct geometric mapping coefficients to complete the quantitative conversion of disturbance to acceleration. By inverting the sign, the disturbance is canceled in the opposite direction. After generating the acceleration correction vector, the target control compensation vector is further obtained. This method can accurately match the mechanical relationship of the hoisting system. The compensation direction is completely opposite to the disturbance direction. At the same time, relying on the real-time updated rope length parameters, the mapping coefficients are dynamically adapted to the working conditions, effectively avoiding the problems of compensation amplitude mismatch and compensation lag, improving the accuracy of load swing cancellation, and enhancing the control stability under complex wind disturbance and rope length variation scenarios.

[0014] In one optional implementation, the target control compensation vector is obtained based on the acceleration correction vector, including: Obtain the initial control variables corresponding to the target UAV; the initial control variables include the initial lateral control variables and the initial longitudinal control variables. Obtain the real-time heading angle of the target drone; Based on the real-time heading angle, a rotation matrix is ​​constructed from the body coordinate system to the geographic Cartesian coordinate system; Based on the rotation matrix, the initial control quantity is transformed to obtain the transformed control quantity in the geographic Cartesian coordinate system; The lateral control compensation vector is obtained by correcting the lateral transfer control quantity in the transfer control quantity based on the lateral correction disturbance acceleration in the acceleration correction vector. Based on the longitudinal correction disturbance acceleration in the acceleration correction vector, the longitudinal conversion control quantity in the conversion control quantity is corrected to obtain the longitudinal control compensation vector; Based on the lateral control compensation vector and the longitudinal control compensation vector, the synthetic control compensation vector is obtained; The synthesized control compensation vector is smoothed and filtered to obtain the target control compensation vector.

[0015] This method constructs a coordinate rotation matrix using real-time heading angles, uniformly transforming the initial control quantities in the body coordinate system to the geographic Cartesian coordinate system, thus eliminating the coupling deviation of control components caused by changes in nose orientation. The method obtains a synthetic compensation vector by superimposing the lateral and longitudinal acceleration correction quantities, and then filters out high-frequency noise introduced by image recognition and observer recursion through smoothing filtering. Finally, it outputs a stable target control compensation vector without abrupt changes, avoiding attitude jitter caused by control command jumps, ensuring that the compensation command matches the global spatial reference, and improving the continuity of trajectory correction and flight stability.

[0016] In one optional implementation, the initial trajectory corresponding to the target UAV is corrected based on the target control compensation vector to obtain the target trajectory corresponding to the target UAV, including: Transform the target control compensation vector into the transformed control compensation vector in the body coordinate system; Separate the lateral transfer control compensation vector and the longitudinal transfer control compensation vector from the transfer control compensation vector; Based on the lateral conversion control compensation vector and the longitudinal conversion control compensation vector, the lateral compensation component and the longitudinal compensation component are calculated respectively. Based on the lateral compensation component and the longitudinal compensation component, the target roll angle correction and the target pitch angle correction are calculated respectively. Obtain the original expected roll angle and original expected pitch angle corresponding to the initial trajectory; Add the target roll angle correction to the original desired roll angle to obtain the corrected desired roll angle; Add the target pitch angle correction to the original desired pitch angle to obtain the corrected desired pitch angle; Obtain the current roll angle and current pitch angle of the target drone; Calculate the roll channel attitude deviation between the corrected desired roll angle and the current roll angle; Calculate the pitch channel attitude deviation between the corrected desired pitch angle and the current pitch angle; Based on the attitude deviations of the roll channel and pitch channel, attitude adjustment commands are generated. Determine the compensation acceleration based on the attitude adjustment command; The initial trajectory is corrected based on the compensation acceleration to obtain the target trajectory corresponding to the target UAV.

[0017] This method transforms the compensation vector from the global coordinate system back to the body coordinate system and decomposes it to generate roll and pitch angle corrections, which are then superimposed on the original desired attitude angle. Combined with real-time attitude feedback, a closed-loop attitude adjustment command is formed, outputting a compensation acceleration adapted to the aircraft's dynamic characteristics. The original flight trajectory is then dynamically corrected. This achieves a complete closed-loop mapping of control compensation from spatial acceleration to UAV attitude and trajectory, taking into account both global disturbance cancellation and onboard attitude stability. It can continuously and smoothly correct the flight path, effectively suppressing trajectory deviations caused by load swings, avoiding sudden attitude changes and flight path oscillations, and significantly improving the trajectory tracking accuracy and flight safety of UAV hoisting operations.

[0018] In a second aspect, the present invention provides a swing suppression device for unmanned aerial vehicle (UAV) hoisting and transportation, the device comprising: The first acquisition module is used to acquire flight payload images corresponding to the target payload of the target UAV during flight. The second acquisition module is used to acquire a reference load image corresponding to the target load of the target UAV in a stable hovering state. The calculation module is used to calculate the target control compensation vector corresponding to the target load of the target UAV during flight, based on the flight load image and the reference load image. The correction module is used to correct the initial trajectory of the target UAV based on the target control compensation vector, so as to obtain the target trajectory of the target UAV.

[0019] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the UAV hoisting and transportation sway suppression method described in the first aspect or any corresponding embodiment thereof.

[0020] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the UAV hoisting and transport sway suppression method of the first aspect or any corresponding embodiment thereof.

[0021] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the UAV hoisting and transportation sway suppression method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic flowchart of the first method for suppressing swaying during unmanned aerial vehicle (UAV) hoisting and transportation according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the second process of the unmanned aerial vehicle (UAV) hoisting and transportation swing suppression method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the third process of the unmanned aerial vehicle (UAV) hoisting and transportation swing suppression method according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the fourth process of the unmanned aerial vehicle (UAV) hoisting and transportation swing suppression method according to an embodiment of the present invention; Figure 6 This is a structural block diagram of a drone hoisting and transport sway suppression device according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] As an optional application scenario of this invention, such as Figure 1 As shown, the drone hoisting and transportation sway suppression system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0028] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0029] According to an embodiment of the present invention, a method for suppressing swing during unmanned aerial vehicle (UAV) hoisting and transportation is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0030] This embodiment provides a method for suppressing sway during drone hoisting and transportation, which can be used in the aforementioned electronic equipment. Figure 2 This is a schematic flowchart of the first type of unmanned aerial vehicle (UAV) hoisting and transportation sway suppression method according to an embodiment of the present invention, as shown below. Figure 2 As shown, the process includes the following steps: Step S201: Obtain the flight payload image corresponding to the target payload of the target UAV during flight.

[0031] The target drone refers to an industrial drone equipped with a vision camera, performing tasks such as hoisting and transporting materials for emergency repairs. The entire sway suppression algorithm runs on the onboard device. For example, it could be a multi-rotor heavy-duty drone equipped with a downward-looking monocular camera, with rescue supplies suspended below by ropes.

[0032] The target payload refers to various goods suspended from the lower end of the drone's cable, whose swing amplitude needs to be monitored visually. Examples include emergency rescue kits, power maintenance tool kits, and small engineering components.

[0033] Flight payload images refer to raw image material captured in real-time by the onboard camera during the dynamic flight phase of the drone, with the entire image containing the suspended payload. This serves as the foundational visual data for subsequent calculations of swing angles and rope lengths. For example, this could be RGB color images of the suspended first-aid kit continuously captured by the onboard downward-looking camera as the drone moves forward.

[0034] Image acquisition refers to the process of capturing images in real time using onboard camera hardware, followed by onboard processor reading and caching the complete image data to complete image acquisition and local storage. For example, the onboard camera continuously captures images of the suspended cargo below at a fixed frame rate, and the flight control chip reads each frame in real time and caches it in memory to complete the acquisition of the flight payload image.

[0035] In this step, the UAV uses its onboard downward-facing monocular camera to capture real-time images of the target payload suspended below, directly generating an RGB color payload image with color information. Simultaneously, to reduce subsequent visual computation, the color image is converted into a single-channel grayscale image that retains only brightness and darkness information. A sliding window, pre-sized according to noise reduction requirements, is used to perform weighted smoothing filtering on the grayscale image. During filtering, the pixel at the center of the window is used as a reference; pixels closer to the center and with smaller differences in brightness and darkness are assigned higher computational weights, having a greater impact on the smoothing result of the center pixel. The sliding window traverses the entire grayscale image row by row and column by column, repeatedly performing weighted calculations across all regions, ultimately removing image noise, preserving the payload outline, and outputting a clean, smooth grayscale image. For example, a drone carrying a first-aid kit flies forward, its onboard camera continuously capturing color images of the kit. After being converted to grayscale, these images are filtered region-by-region using a 3x3 sliding window. Pixels with similar grayscale values ​​inside the kit have higher weights, while background noise pixels have lower weights. After the window traverses the entire image, a clear, noise-free grayscale image is obtained, facilitating subsequent identification of the load area and calculation of sway parameters. It should be noted that the sliding window is set based on the local pixel neighborhood range required for image filtering.

[0036] Step S202: Obtain the reference load image corresponding to the target payload when the target UAV is in a stable hovering state.

[0037] Stable hovering refers to a state in which a drone remains in a fixed position in the air, with no deviation or swaying of the fuselage, and the load suspended below hangs naturally and vertically, and the whole drone is in a static and stable state. For example, when a drone lifts goods into the air and stays still in place, with the rope vertical and the goods not swaying from side to side, this is a stable hovering state.

[0038] A reference payload image refers to a standard reference image taken when the UAV is hovering and the payload is vertical and without swaying. It is used to calibrate the standard pixel area and reference rope length, serving as a comparison reference for subsequent dynamic flight images. For example, the reference payload image can be a standard view of the toolbox being hoisted while hovering, which is then used to calculate the standard reference pixel area.

[0039] For details regarding the target payload, target UAV, and acquisition, please refer to the above text, which will not be repeated here.

[0040] In this step, a standard reference image is acquired under stable conditions where the drone is completely stationary and hovering, and the load is vertical and without swaying. This reference image serves as the core reference for subsequent calculations of the dynamic rope length and is used in conjunction with dynamic load images captured during flight to calculate the real-time rope length and sway angle. In actual operation, after the drone is loaded with relief supplies and takes off, it first hovers at a fixed point until the cargo is completely stable and the rope is vertical and without deviation. The onboard downward-facing camera then captures a complete image containing the supplies and caches it as the reference load image.

[0041] Step S203: Based on the flight load image and the reference load image, calculate the target control compensation vector corresponding to the target load of the target UAV during flight.

[0042] The target control compensation vector refers to a set of vector-based control corrections used to counteract flight disturbances caused by load fluctuations and is the core basis for adjusting the flight path. For example, it contains vector data with two sets of acceleration correction values, one lateral and one longitudinal. There are no specific limitations on its type; it can be set according to actual needs.

[0043] The calculation process involves using the visual information carried by two types of images as input data, processing it through a complete set of control algorithms, and directly outputting a compensation vector result that can be used for trajectory correction. Various parameter derivations and estimations in between are all part of the internal calculation process. For example, by simultaneously reading the hovering reference image and the real-time flight image, the entire flight control backend completes the calculations and ultimately outputs the target control compensation vector.

[0044] This step uses the flight load image and the reference load image as input materials to obtain the target control compensation vector. The internal calculations include visual recognition, swing angle rope length calculation, and disturbance observation conversion. During operation, the flight control system simultaneously retrieves the two previously acquired images. Based on the comparison of load contours, pixel sizes, and positional offsets in the two images, it comprehensively calculates all the interference caused by the current load and uniformly converts it into the target control compensation vector. This vector will be directly used in subsequent trajectory correction steps. In a specific embodiment, a mountain power inspection UAV is carrying a toolbox. During the hovering phase, a reference load image is retained. During forward flight, flight load images are continuously acquired. The flight control system directly uses these two types of images as input to complete the overall calculation and outputs the corresponding target control compensation vector at once. Subsequently, the preset flight path is corrected based on this vector to counteract the forward and backward, left and right swaying of the toolbox caused by strong winds.

[0045] Step S204: Based on the target control compensation vector, correct the initial trajectory corresponding to the target UAV to obtain the target trajectory corresponding to the target UAV.

[0046] The initial trajectory can be a standard flight path pre-planned before the drone operation, without considering load swing interference, only meeting the basic transportation point requirements. For example, a preset flight path from a material storage point to a rescue point in a straight line, without any swing correction actions.

[0047] The target trajectory refers to the newly updated flight path obtained after superimposed compensation and correction. It can actively counteract load fluctuations, ensuring stable flight and preventing deviation from the flight path. For example, after superimposed sway suppression correction for a UAV, a smooth and stable transport route is obtained, and the cargo will not sway significantly during flight.

[0048] In a specific embodiment, in a mountainous disaster relief scenario, a heavy-duty drone carrying a first-aid kit is planned with a straight initial trajectory leading directly to the rescue point before takeoff. After image processing in the above steps, a target control compensation vector is obtained to counteract the left-right swaying of the kit caused by strong winds. When executing this step, the flight controller makes an overall correction to the initial trajectory based on the target control compensation vector and fine-tunes the flight path in real time, ultimately generating a target trajectory with minimal swaying of the kit throughout the entire journey and without deviating from the work area. The drone completes the material transportation according to this target trajectory.

[0049] The UAV hoisting and transportation sway suppression method provided in this embodiment generates a compensation vector to correct the flight trajectory through two types of load image calculations. It does not require additional sensors, can adapt to changing hoisting conditions, suppress load sway in real time, reduce hardware costs, and improve flight stability and safety.

[0050] This embodiment provides a method for suppressing swing during unmanned aerial vehicle (UAV) hoisting and transportation, which can be used on the aforementioned mobile terminals, such as mobile phones and tablets. Figure 3 This is a flowchart of a method for suppressing swaying during unmanned aerial vehicle (UAV) hoisting and transportation according to an embodiment of the present invention, as shown below. Figure 3 As shown, the process includes the following steps: Step S301: Obtain the flight payload image corresponding to the target payload of the target UAV during flight. For details, please refer to [link to details]. Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0051] Step S302: Obtain the reference load image corresponding to the target payload of the target UAV in a stable hovering state. For details, please refer to [link to relevant documentation]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0052] Step S303: Based on the flight load image and the reference load image, calculate the target control compensation vector corresponding to the target load of the target UAV during flight.

[0053] Specifically, step S303 includes: Step S3031: Perform target recognition on the flight payload image to determine the target payload region and payload outline pixel area corresponding to the target payload in the flight payload image.

[0054] The target load region refers to the connected pixel region within an image that is surrounded by a complete closed contour of the load, representing the imaging range of the cargo. Continuous closed edge lines are identified in a smooth grayscale image, and the connected region surrounded by these continuous closed edge lines is defined as the target load region. For example, gradient magnitude calculation can be performed on the smooth grayscale image to locate pixel abrupt change points. These pixel abrupt change points, which meet geometric closure requirements, can be connected to form continuous closed edge lines. Internal pixel filling is then performed on these continuous closed edge lines to obtain a connected region, which is then defined as the target load region.

[0055] The load outline pixel area refers to the total number of pixels contained within the target load region, used to determine the real-time length of the suspension rope. The geometric centroid coordinates of the target load region are used as the load pixel position coordinates, and the total number of pixels contained within the target load region is used as the load outline pixel area. For example, the first and zeroth moments of the target load region can be calculated, and the ratio of the first to zeroth moments can be used to determine the geometric centroid coordinates, which are then used as the load pixel position coordinates. The total number of discrete pixel units contained within the target load region is then counted, and this total number of discrete pixel units is used as the load outline pixel area. No specific implementation is limited here; the specific implementation can be set according to actual needs.

[0056] It should be noted that the geometric centroid coordinates of the load target area refer to the arithmetic mean of the horizontal and vertical coordinates of all pixels in the area, representing the geometric center of the load shape in the load image.

[0057] Step S3032: Based on the target load area, calculate the real-time load swing angle corresponding to the target load of the target UAV during flight.

[0058] The real-time swing angle of the load includes both the real-time swing angle in the lateral direction and the real-time swing angle in the longitudinal direction. The UAV suspends the load via a single cable, forming a two-dimensional pendulum system. The suspended load does not swing in only one direction; it simultaneously generates two sets of independent swing motions, corresponding to two independent offsets. Therefore, these are calculated and controlled separately for the lateral and longitudinal angles. Specifically, the load's swing is divided into two sets of mutually perpendicular movements: a longitudinal swing (forward and backward) and a lateral swing (left and right). These two sets of movements do not interfere with each other. The longitudinal swing occurs along the forward and backward plane from the UAV's nose to its tail, triggered by acceleration, deceleration, and pitch, and only affects the pitch control channel. The lateral swing occurs along the left and right side flight planes of the fuselage, triggered by crosswinds and lateral movement.

[0059] Meanwhile, the camera image itself has mutually perpendicular horizontal X-axis and vertical Y-axis. The pixel offset of the load on the X-axis corresponds to the left and right sway of the body, and the offset on the Y-axis corresponds to the back and forth sway. By offset in the X-axis and Y-axis directions, the swaying of the drone in the horizontal and vertical directions can be located.

[0060] In this step, the cargo imaging area defined in the image is used as the basis for calculation. The offset information of the cargo relative to the center of the camera image is extracted from it, and the current swing and tilt degree of the hoisted cargo is calculated. For example, when the cargo is in the wind or when the drone changes speed, it will swing in two vertical directions at the same time: forward and backward, and left and right. After the camera images the cargo, the two types of swaying will be reflected in the horizontal and vertical pixel offsets of the image, respectively. The tilt angles in the two directions are calculated separately, and then integrated to obtain the complete real-time swing angle of the load.

[0061] Step S3033: Perform target recognition on the reference load image to determine the reference load pixel area and reference physical rope length corresponding to the reference load image.

[0062] Static calibration is performed while the drone remains hovering in a fixed position and the cargo hangs vertically without swaying along the sling. This yields two standard parameters: the reference load pixel area and the reference physical sling length. Specifically, the onboard image sensor captures an image of the load for calibration. The Canny algorithm is then used to denoise and smooth the image. The strength and orientation of grayscale changes in each pixel are calculated. Edge lines are refined using non-maximum suppression, and false contours are removed using double thresholding and edge connectivity techniques, preserving the true boundaries of the cargo. The complete closed contour of the cargo is then extracted, and the total number of pixels within the contour is counted; this value serves as the reference load pixel area. With the drone hovering stably and the load hanging vertically, the physical vertical distance from the drone's bottom attachment point to the load's center of mass is measured in real-time using an onboard laser rangefinder to obtain the reference physical sling length. The implementation method is not limited; it can be set according to actual needs.

[0063] Step S3034: Based on the payload outline pixel area, the reference payload pixel area, and the reference physical rope length, calculate the dynamic rope length corresponding to the target payload during the flight of the target UAV.

[0064] The dynamic rope length parameter reflects the current change in the suspension length between the UAV and the payload, characterizing the real-time changes in rope length during flight as the payload's position, attitude, and motion change. The dynamic rope length parameter is output by calculating the square root ratio of the reference pixel area to the payload's outer contour pixel area, and then multiplying this square root ratio by the reference physical rope length. For example, the reference pixel area and the payload's outer contour pixel area are divided to obtain the area ratio; the square root of this area ratio is then calculated to obtain the square root ratio; finally, this square root ratio is multiplied by the reference physical rope length to output the dynamic rope length parameter.

[0065] Step S3035: Calculate the target control compensation vector based on the real-time swing angle of the load and the dynamic rope length.

[0066] The real-time swing angle of the load is divided into two independent sets of angles: lateral and longitudinal, representing the degree to which the cargo tilts relative to the vertical line in four directions: left and right, forward and backward. Under the same wind force, the larger the swing angle value, the more violent the cargo swaying. This serves as the basis for judging the strength of the disturbance. The two angles are independent of each other and correspond to the roll and pitch control channels of the UAV, respectively.

[0067] The target control compensation vector is a two-dimensional vector consisting of two sets of correction components, one horizontal and one vertical. The two components correspond to the reverse control quantities required to counteract left-right and forward-backward swaying, respectively. The vector carries its own magnitude and direction information and is directly transmitted to the flight control system as the basis for attitude and flight path adjustments.

[0068] The disturbance force generated by the load is affected by both the swing angle and the length of the hoisting rope. Relying solely on the swing angle ignores the mechanical differences caused by changes in rope length, leading to significant deviations in the compensation effect. Therefore, the calculations separate and treat the forward / backward and left / right directions independently. Specifically, for longitudinal swing, the longitudinal correction amount to counteract forward / backward sway is calculated based on the current dynamic rope length; for lateral swing, the lateral correction amount to counteract left / right sway is calculated based on the rope length at the same moment. Calculating the two directions separately avoids interference between the two types of sway and prevents confusion in the compensation direction. After the two sets of correction values ​​are integrated into a target control compensation vector, they are uniformly transmitted to the subsequent trajectory correction stage. The UAV adjusts its attitude in real time according to the vector to actively counteract the flight deviation caused by cargo sway, reducing the risk of overall aircraft vibration.

[0069] Step S304: Based on the target control compensation vector, correct the initial trajectory corresponding to the target UAV to obtain the target trajectory corresponding to the target UAV.

[0070] Please see details Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0071] The UAV hoisting and transportation sway suppression method provided in this embodiment acquires all parameters through an airborne camera, eliminating the need for additional sensors such as laser rangefinders and external inertial sensors, thus reducing the UAV's load and lowering hardware assembly and calibration costs. It statically calibrates the standard pixel area and actual rope length using a hovering reference image, then dynamically calculates the real-time rope length by comparing the load pixel size with the flight image. Combined with the load area center offset, it calculates the lateral and longitudinal sway angles, fully matching the mechanical characteristics of a single pendulum during hoisting. This allows it to adapt to various working conditions such as flight altitude, rope extension and retraction, and load size changes. Simultaneously, it uses the real-time sway angle and dynamic rope length to jointly solve for the control compensation vector, distinguishing between front-to-back and left-to-right disturbances to generate corresponding correction amounts, avoiding mutual coupling interference between swaying in different directions. This accurately counteracts load swaying caused by gusts and acceleration / deceleration. The entire process involves airborne closed-loop real-time calculation with fast compensation response, significantly reducing the risk of flight oscillations and large cargo swaying. It is suitable for various UAV hoisting scenarios such as logistics, inspection, and disaster relief, exhibiting greater versatility and adaptability to complex field environments.

[0072] Figure 4 This is a flowchart illustrating the calculation of the real-time swing angle of the load in the unmanned aerial vehicle (UAV) hoisting and transportation swing suppression method according to an embodiment of the present invention. Figure 4 As shown, step S3032 above further includes the following steps: Step S401: Obtain the center coordinates of the payload image corresponding to the flight payload image.

[0073] The load image center coordinates refer to the coordinates obtained with the top left corner of the image as the origin. The horizontal and vertical pixel values ​​of the image center point, calculated based on the total horizontal and vertical pixel dimensions of the image, serve as a fixed reference benchmark for judging cargo offset.

[0074] The center coordinates of the load image are determined based on its resolution, and a two-dimensional pixel deviation vector between the load pixel position coordinates and the center coordinates of the load image is calculated. For example, the horizontal and vertical resolutions of the load image are read, and half of the horizontal resolution and half of the vertical resolution are combined to obtain the center coordinates of the load image.

[0075] Step S402: Based on the target load area, determine the load pixel position coordinates corresponding to the geometric center of the target load.

[0076] The geometric centroid coordinates of the load target region are used as the load pixel position coordinates. The first moment and zero moment of the load target region are calculated. The geometric centroid coordinates are determined by the ratio of the first moment to the zero moment. The geometric centroid coordinates are used as the load pixel position coordinates. The geometric centroid coordinates of the load target region refer to the arithmetic mean of the horizontal and vertical coordinates of all pixels in the region, which represents the geometric center of the load shape in the load image.

[0077] Step S403: Based on the load pixel position coordinates and the load image center coordinates, calculate the two-dimensional pixel deviation vector.

[0078] The two-dimensional pixel deviation vector is a two-dimensional vector composed of the difference in horizontal coordinates and the difference in vertical coordinates. The two components represent the offset distance and offset direction of the cargo relative to the center of the image, respectively, and are the basic data for calculating the true swing angle.

[0079] The pixel deviation vector is obtained by performing a difference operation between the load pixel position coordinates and the load image center coordinates; the pixel deviation vector is then subjected to a coordinate axis rotation transformation to obtain a two-dimensional pixel deviation vector.

[0080] Step S404: Based on the lateral deviation vector in the two-dimensional pixel deviation vector, calculate the real-time lateral swing angle corresponding to the target payload of the target UAV during flight.

[0081] The two-dimensional pixel deviation vector is converted into the real-time swing angle of the load in three-dimensional space through inverse trigonometric geometry calculation. For example, the intrinsic parameter matrix of the airborne monocular camera is queried, and the focal length parameter in the intrinsic parameter matrix is ​​extracted; the lateral component of the two-dimensional pixel deviation vector is divided by the focal length parameter to obtain a set of angle tangent values, and the arctangent function is calculated on the set of angle tangent values. Through the above inverse trigonometric geometry calculation process, the real-time swing angle of the load is obtained.

[0082] The real-time swing angle of the load can be expressed as: .

[0083] in, This indicates the real-time swing angle of the load in the lateral direction; This represents the horizontal component of the two-dimensional pixel offset vector. This indicates the focal length parameter of the airborne monocular camera.

[0084] Step S405: Based on the longitudinal deviation vector in the two-dimensional pixel deviation vector, calculate the longitudinal real-time swing angle corresponding to the target payload of the target UAV during flight.

[0085] The two-dimensional pixel deviation vector is converted into the real-time swing angle of the load in three-dimensional space through inverse trigonometric geometry calculation. For example, the intrinsic parameter matrix of the airborne monocular camera is queried, and the focal length parameter in the intrinsic parameter matrix is ​​extracted; the vertical component of the two-dimensional pixel deviation vector is divided by the focal length parameter to obtain a set of angle tangent values, and the arctangent function is calculated on the set of angle tangent values. Through the above inverse trigonometric geometry calculation process, the real-time swing angle of the load is obtained.

[0086] The real-time swing angle of the load can be expressed as: .

[0087] in, This indicates the real-time swing angle of the load in the longitudinal direction; This represents the vertical component of the two-dimensional pixel offset vector. This indicates the focal length parameter of the airborne monocular camera.

[0088] Step S406: Calculate the real-time swing angle of the load based on the real-time lateral swing angle and the real-time longitudinal swing angle.

[0089] Lateral and longitudinal oscillations are two independent motions that are perpendicular to each other and uncoupled. Neither angle alone can fully describe the overall swaying state of the load. Therefore, the sway angles of the two groups are vector-synthesized to obtain a unified comprehensive sway parameter, which is the real-time sway angle of the load. This allows for an intuitive assessment of the total swaying intensity of the load, and the unified data format can be directly input into the subsequent extended state observer to construct an augmented state vector, providing a complete sway angle input for disturbance estimation and compensation vector solution.

[0090] The UAV hoisting and transportation sway suppression method provided in this embodiment extracts the geometric center coordinates of the load based on the image center, calculates the pixel deviation independently in the horizontal and vertical directions, and solves the corresponding sway angles respectively, thereby distinguishing between the two types of sway. It can optimize the data processing logic in each direction and then synthesize the overall sway angle to provide a complete input for disturbance observation. The step-by-step operation also facilitates the individual correction of abnormal data, effectively improving the accuracy of sway angle calculation and the adaptability of the algorithm to the environment, and preventing the compensation control direction from being disordered and the sway suppression from failing.

[0091] Figure 5 This is a flowchart illustrating the calculation of the target control compensation vector in the UAV hoisting and transportation sway suppression method according to an embodiment of the present invention. Figure 5 As shown, step S3035 above further includes the following steps: Step S501: Calculate the load swing angular velocity corresponding to the target load based on the real-time swing angle of the load.

[0092] Load oscillation angular velocity refers to the change in load oscillation angle per unit time. It reflects the speed of load swaying and the severity of disturbance changes, and is a key parameter for predicting swaying trends and implementing compensatory control in advance. The load oscillation angular velocity is obtained by performing first-order time-delay differential processing on the real-time load oscillation angle. For example, the real-time load oscillation angle is discretized and sampled, storing the real-time load oscillation angle at adjacent sampling times; the difference between the real-time load oscillation angle at the current sampling time and the real-time load oscillation angle at the previous sampling time is calculated to obtain the angle increment; the angle increment is divided by the sampling period, and the result of the division is subjected to first-order low-pass filtering to obtain the load oscillation angular velocity.

[0093] Step S502: Based on the real-time swing angle of the load, the dynamic rope length, and the angular velocity of the load swing, an augmented state vector is generated.

[0094] Augmented state vectors are multidimensional state arrays formed by expanding the total quantities of unmodeled dynamics and external disturbances that cannot be directly measured into new states based on the original observable motion state variables of the system. They are used to expand the state observer to simultaneously estimate the system's operating state and comprehensive disturbances.

[0095] The dynamic rope length parameters, real-time load swing angle, and load swing angular velocity are time-domain aligned and vector-stitched to generate an augmented state vector. For example, using a unified sampling clock as the time-domain reference, the dynamic rope length parameters, real-time load swing angle, and load swing angular velocity acquired at the current moment are synchronously acquired and timestamped. The sampling times of each parameter are checked for consistency. If the sampling times are consistent, the corresponding values ​​are directly read. If there is a discrepancy in the sampling times of each parameter, interpolation compensation, nearest-time matching, or delay-and-hold processing are performed on the asynchronous parameters to complete the time-domain alignment. After time-domain alignment, the dimensions are arranged and vector-stitched according to a preset topological order to generate the augmented state vector.

[0096] It should be noted that the topological order is set based on the arrangement logic of the physical meaning of each dimension of the augmented state vector.

[0097] Step S503: Configure the target gain matrix of the extended state observer according to the dimension of the augmented state vector.

[0098] For example, the dimension of the augmented state vector is read, and a one-to-one correspondence between each state component and the observation error is established according to the arrangement order of the dynamic rope length parameter, the real-time swing angle of the load, and the angular velocity of the load swing in the vector. Based on the established correspondence, the intensity of the observation error on different state components is distinguished by the rate of change and the degree of mutual influence of each state component during the load swing process. State components with faster change rates correspond to higher correction intensities, and state components with slower change rates correspond to gentler correction intensities. The influence range of the observation error between the current sampling time and the adjacent sampling time is constrained based on the correction intensity, so that the error correction process remains continuous in the time dimension. The desired convergence characteristics are set according to the pole placement method. By establishing a correspondence between the desired convergence speed and the correction intensity of each state component, the correction weights of each state component are adjusted, and the correction intensities corresponding to each state component are arranged according to the dimension order of the augmented state vector to generate a gain matrix.

[0099] It should be noted that in the extended state observer, the gain matrix is ​​used to adjust the weight set of the internal state correction intensity of the observation error, which determines the speed and smoothness of the convergence of the observation results to the true value. The extended state observer is set based on the synchronous sampling step size of the payload image acquisition period and the UAV flight control frequency. For example, when there is a difference between the control refresh period corresponding to the payload image acquisition period and the UAV flight control frequency, the unified time resolution between the two is used as the discrete update step size of the extended state observer, so that the extended state observer receives the augmented state vector corresponding to the current time in each sampling step and completes one state recursion and perturbation estimation.

[0100] Pole placement is a design method based on state-space analysis in control theory. It is used to change the dynamic characteristics of a controlled object and enable it to achieve the expected performance indicators. The principle is to move the poles in the closed-loop state to a specified position on the complex plane by introducing state feedback.

[0101] Step S504: Input the augmented state vector into the extended state observer and perform discrete-time recursion based on the target gain matrix to output the total disturbance estimate.

[0102] For example, the augmented state vector at the current sampling moment is input into the extended state observer, and the internal state value stored at the previous sampling moment is read. The current observation error is calculated based on the difference between the two. According to the weight relationship of each dimension in the gain matrix, the observation error is weighted and allocated to each component of the dynamic rope length parameter, the real-time swing angle of the load, and the angular velocity of the load swing, respectively, to obtain the correction amount used to correct the internal state change. In the state recursion process, with the current internal state as the initial value, multiple intermediate moments are selected within a sampling step, and the state change trend at each intermediate moment is estimated successively. The multiple intermediate change trends are combined and calculated according to the weighting rule of the Runge-Kutta method to obtain the state update value at the end of the sampling step. The correction amount is superimposed on each state component corresponding to the state update value, so that the updated internal state simultaneously includes the original state evolution trend and the observation error correction effect. In the updated internal state vector, the components related to the disturbance in the extended dimension are extracted as the total disturbance estimate according to the arrangement order of the augmented state vector.

[0103] It should be noted that the total disturbance estimate is an estimate of the comprehensive influencing factors during the load swing process. It is used to provide a basis for subsequent compensation and correction, so that the control process can respond in a timely manner to the swing trend, inertial influence and external disturbances, thereby reducing the adverse effects of load swing on flight attitude and track maintenance.

[0104] Runge-Kutta method is a discretization method used in numerical analysis to solve initial value problems of ordinary differential equations. It is used to obtain approximate values ​​of numerical solutions to differential equations through iterative calculation. The principle is to construct a numerical integration formula with high order accuracy by weighted averaging of the slopes of multiple predicted points within each step.

[0105] Step S505: Calculate the target control compensation vector based on the total disturbance estimate.

[0106] The total disturbance estimate is used to quantify the magnitude of all disturbances that cause load swing. Based on this, the control quantity required to counteract the disturbance is calculated in reverse. The correction components are calculated separately for the front and rear, and left and right independent channels, and then combined to obtain the target control compensation vector.

[0107] By integrating the real-time swing angle of the load, the dynamic rope length, and the angular velocity of the load swing to construct an augmented state vector, and matching the observer gain matrix according to the vector dimension, the algorithm dimension matching is ensured to avoid computational anomalies. The comprehensive total disturbance is estimated in real time through recursion by the extended state observer. Various unknown disturbances such as wind disturbance and variable speed inertia can be quantified without additional sensors. Then, the control compensation vector is solved based on the disturbance value to achieve active sway suppression. The integrated observation of swing state, rope length parameters, and external disturbances is fully realized, and the collaborative operation improves the disturbance estimation accuracy.

[0108] Equivalent disturbance components are extracted from the total disturbance estimate. For example, disturbance components corresponding to the horizontal transverse axis and the vertical axis are extracted from the total disturbance estimate. The horizontal transverse axis disturbance components and the vertical axis disturbance components are combined with the dynamic rope length parameter and gravitational acceleration constant at the current moment for dimensional conversion. Based on the small-angle geometric relationship of the load swing, the converted horizontal transverse axis disturbance components and the vertical axis disturbance components are decomposed in direction and converted numerically to correct the amplitude deviation caused by the inconsistency of the scales of different state components, so as to obtain the transverse equivalent disturbance components and the vertical equivalent disturbance components.

[0109] Furthermore, the gravitational acceleration constant and the dynamic rope length parameter are divided to obtain a ratio; the square root of the ratio is then performed to obtain the geometric mapping coefficient. The gravitational acceleration constant acquired by the airborne sensor is divided by the current dynamic rope length parameter to establish a geometric proportional relationship between the load swing arc and the horizontal displacement. The geometric mapping coefficient is multiplied by the lateral equivalent disturbance component and the longitudinal equivalent disturbance component respectively to obtain the lateral original disturbance acceleration and the longitudinal original disturbance acceleration; the signs of the lateral original disturbance acceleration and the longitudinal original disturbance acceleration are then inverted to obtain the lateral corrected disturbance acceleration and the longitudinal corrected disturbance acceleration; finally, a preset constraint threshold is used to numerically constrain the reverse acceleration to generate an acceleration correction vector.

[0110] It should be noted that the constraint threshold is set based on the physical output limit of the actuator, such as the drone motor. For example, the range of values ​​can be... The basis for the example value range is: since the power output of the actuator has saturation characteristics, if the value of the acceleration correction vector exceeds the adjustment capability of the actuator, it will cause the control command to fail or cause the UAV to lose attitude control. Therefore, it is necessary to set this range to ensure the physical feasibility of the compensation amount.

[0111] For example, the initial control quantities output by the UAV under the current mission trajectory are read from the flight control bus, and the initial lateral control quantity and initial longitudinal control quantity are extracted from the initial control quantities; the heading angle information of the UAV at the current moment is read simultaneously, and the azimuth deflection relationship of the body coordinate system relative to the geographic reference coordinate system is determined based on the heading angle information. A rotation matrix from the body coordinate system to the geographic Cartesian coordinate system is constructed. For example, based on the azimuth deflection relationship, a two-dimensional coordinate rotation relationship corresponding to the current heading angle is constructed. The initial lateral control quantity and initial longitudinal control quantity in the initial control quantities are decomposed and recombined according to the two-dimensional coordinate rotation relationship, so that the control quantity originally defined based on the current nose direction of the UAV is converted into axial control components defined based on the Cartesian coordinate system. The converted axial control components are uniformly arranged to obtain the converted control quantity consistent with the acceleration correction vector coordinate reference.

[0112] It should be noted that the initial control quantity refers to the original control command output by the flight control bus when the UAV is executing the current mission trajectory. It usually corresponds to the horizontal and longitudinal control components and is the initial control intention without considering load swing compensation.

[0113] Transformed control quantities refer to the control quantities that are transformed from the body coordinate system, such as the coordinate system set based on the nose direction, to the geographic reference coordinate system, such as the Cartesian coordinate system, through a yaw angle rotation, so that they are consistent with the coordinate reference of the acceleration correction vector.

[0114] The lateral control compensation vector is obtained by correcting the lateral conversion control quantity in the conversion control quantity using the lateral correction disturbance acceleration in the acceleration correction vector; the longitudinal control compensation vector is obtained by correcting the longitudinal conversion control quantity in the conversion control quantity using the longitudinal correction disturbance acceleration in the acceleration correction vector; and a composite control compensation vector is obtained based on the obtained lateral and longitudinal control compensation vectors. For example, the components of the acceleration correction vector and the conversion control quantity along the lateral and longitudinal axes are extracted, and their components are added to obtain the lateral and longitudinal control compensation vectors. The lateral and longitudinal control compensation vectors are then vector-joined to obtain the composite control compensation vector. The composite control compensation vector is input to a preset low-pass filter. The low-pass filter performs high-frequency noise filtering and signal smoothing operations on the composite control compensation vector to form the target control compensation vector.

[0115] It should be noted that the low-pass filter is set based on the frequency characteristics of the noise to be filtered out and the signal smoothness requirements.

[0116] The target control compensation vector is transformed into a transformation control compensation vector in the body coordinate system. The lateral and longitudinal transformation control compensation vectors are then separated from the transformation control compensation vector. Orthogonal decomposition is performed on the target control compensation vector in the body coordinate system to obtain the lateral and longitudinal compensation components. The lateral and longitudinal compensation components are then divided by the gravitational acceleration constant to obtain the angle tangent mapping value. An arctangent operation is performed on the angle tangent mapping value to obtain the target roll angle correction and the target pitch angle correction.

[0117] The system reads the original desired roll angle and original desired pitch angle from the attitude control loop, adds the target roll angle correction to the original desired roll angle to obtain the corrected desired roll angle, and adds the target pitch angle correction to the original desired pitch angle to obtain the corrected desired pitch angle. It then obtains the current roll angle and current pitch angle corresponding to the target UAV, inputs the corrected desired roll angle and corrected desired pitch angle into the attitude controller, calculates the roll channel attitude deviation between the corrected desired roll angle and the current roll angle, and calculates the pitch channel attitude deviation between the corrected desired pitch angle and the current pitch angle. Based on the roll channel attitude deviation and pitch channel attitude deviation, it performs attitude closed-loop adjustment calculations to generate corresponding attitude adjustment commands. These commands are then distributed to the speed control channels of each motor. By changing the output speed of each motor, the overall thrust direction of the UAV is adjusted relative to the body coordinate system. Under the condition of thrust direction tilt, the projection component of the total thrust on the horizontal plane changes, thereby forming compensation accelerations corresponding to the target control compensation vector on the horizontal and vertical axes.

[0118] The compensation acceleration is the horizontal acceleration component generated by the UAV during attitude deflection based on the target control compensation vector. It is used to counteract disturbances caused by load swings and correct flight attitude and trajectory. The attitude control loop is a control loop that performs closed-loop comparison and adjustment of the desired roll angle, desired pitch angle, and current roll angle and current pitch angle, and achieves attitude stabilization and attitude tracking of the UAV by driving changes in the output of each motor.

[0119] By using compensation acceleration to perform real-time vector compensation on the coordinate points of the initial trajectory, the spatial coordinate sequence after offsetting the influence of load swing is calculated, and the target trajectory is generated.

[0120] Furthermore, the compensation acceleration is integrally calculated on the time axis to obtain the position offset vector of the UAV in the three spatial dimensions caused by the load swing. The position offset vector is then vector-summed with the corresponding original coordinate points in the initial trajectory to obtain a spatial coordinate sequence that can counteract the swing interference. Spline interpolation and smoothing are performed on the spatial coordinate sequence to construct the target trajectory, expressed as:

[0121] in, It is the target trajectory; It is the initial trajectory; It is a compensating acceleration; It is the time variable for integration operations.

[0122] It should be noted that the initial trajectory is set based on the UAV mission objectives and obstacle avoidance safety constraints, and is the original ideal spatial coordinate sequence used to guide the UAV's flight.

[0123] The target trajectory is a sequence of spatial coordinates obtained by dynamically adjusting the initial trajectory based on the compensation acceleration. It is used to counteract the effects of disturbances such as load swing on the flight path and improve trajectory tracking accuracy.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the core ideas and scope of protection of the present invention, and all such modifications and equivalent substitutions fall within the scope of protection of the present invention.

[0125] This embodiment also provides a drone hoisting and transportation sway suppression device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0126] This embodiment provides a device for suppressing the swaying of drones during hoisting and transportation, such as... Figure 6 As shown, it includes: The first acquisition module 601 is used to acquire the flight payload image corresponding to the target payload of the target UAV during flight.

[0127] The second acquisition module 602 is used to acquire a reference load image corresponding to the target load of the target UAV in a stable hovering state.

[0128] The calculation module 603 is used to calculate the target control compensation vector corresponding to the target load of the target UAV during flight, based on the flight load image and the reference load image.

[0129] The correction module 604 is used to correct the initial trajectory of the target UAV based on the target control compensation vector, so as to obtain the target trajectory of the target UAV.

[0130] In some alternative implementations, the computing module 603 includes: The first determining unit is used to perform target recognition on the flight payload image and determine the target payload region and the payload outline pixel area corresponding to the target payload in the flight payload image.

[0131] The first calculation unit is used to calculate the real-time swing angle of the target load corresponding to the target load during the flight of the target UAV, based on the target load area.

[0132] The second determining unit is used to perform target recognition on the reference load image and determine the reference load pixel area and reference physical rope length corresponding to the reference load image.

[0133] The second calculation unit calculates the dynamic rope length corresponding to the target payload during the flight of the target UAV, based on the payload outline pixel area, the reference payload pixel area, and the reference physical rope length. The third calculation unit calculates the target control compensation vector based on the real-time swing angle of the load and the dynamic rope length.

[0134] In some alternative implementations, the first computing unit includes: The acquisition submodule is used to obtain the center coordinates of the payload image corresponding to the flight payload image.

[0135] The determination submodule is used to determine the load pixel position coordinates corresponding to the geometric center of the target load based on the target load region.

[0136] The first calculation submodule is used to calculate the two-dimensional pixel deviation vector based on the load pixel position coordinates and the load image center coordinates.

[0137] The second calculation submodule is used to calculate the real-time lateral swing angle of the target payload during the flight of the target UAV based on the lateral deviation vector in the two-dimensional pixel deviation vector.

[0138] The third calculation submodule is used to calculate the real-time longitudinal sway angle of the target payload during the flight of the target UAV based on the longitudinal deviation vector in the two-dimensional pixel deviation vector.

[0139] The fourth calculation submodule is used to calculate the real-time swing angle of the load based on the real-time lateral swing angle and the real-time longitudinal swing angle.

[0140] In some alternative implementations, the third computing unit includes: The angular velocity calculation submodule is used to calculate the load swing angular velocity corresponding to the target load based on the real-time swing angle of the load.

[0141] The generation submodule is used to generate an augmented state vector based on the real-time swing angle of the load, the dynamic rope length, and the angular velocity of the load swing.

[0142] The configuration submodule is used to configure the target gain matrix of the extended state observer based on the dimension of the augmented state vector.

[0143] The output submodule is used to input the augmented state vector into the extended state observer and perform discrete-time recursion based on the target gain matrix to output the total perturbation estimate.

[0144] The target control compensation vector calculation submodule is used to calculate the target control compensation vector based on the total disturbance estimate.

[0145] In some optional implementations, the target control compensation vector calculation submodule includes: The decomposition sub-unit is used to decompose the total disturbance estimate to obtain the lateral equivalent disturbance component and the longitudinal equivalent disturbance component.

[0146] The mapping coefficient calculation subunit is used to calculate the geometric mapping coefficient based on the dynamic rope length and gravitational acceleration.

[0147] The multiplication sub-unit is used to multiply the geometric mapping coefficients with the transverse equivalent disturbance component and the longitudinal equivalent disturbance component, respectively, to obtain the transverse original disturbance acceleration and the longitudinal original disturbance acceleration.

[0148] The sign-inverting sub-unit is used to invert the signs of the original lateral and longitudinal disturbance accelerations to obtain the corrected lateral and longitudinal disturbance accelerations.

[0149] The acceleration correction vector calculation subunit is used to calculate the acceleration correction vector based on the lateral and longitudinal correction perturbation accelerations.

[0150] The sub-unit is obtained, which is used to obtain the target control compensation vector based on the acceleration correction vector.

[0151] In some alternative implementations, obtaining the subunit includes: The first acquisition operator is used to acquire the initial control variables corresponding to the target UAV; the initial control variables include the initial lateral control variables and the initial longitudinal control variables.

[0152] The second acquisition operator is used for the real-time heading angle of the target UAV.

[0153] Construct an operator to generate a rotation matrix from the body coordinate system to the geographic Cartesian coordinate system based on the real-time heading angle.

[0154] The first operator is used to perform coordinate transformation on the initial control quantity based on the rotation matrix, resulting in the transformed control quantity in the geographic Cartesian coordinate system.

[0155] The second operator is used to correct the lateral conversion control quantity in the conversion control quantity based on the lateral correction disturbance acceleration in the acceleration correction vector, thereby obtaining the lateral control compensation vector.

[0156] The third operator is used to correct the longitudinal conversion control quantity in the conversion control quantity based on the longitudinal correction disturbance acceleration in the acceleration correction vector, thereby obtaining the longitudinal control compensation vector.

[0157] The fourth operator is used to obtain the composite control compensation vector based on the lateral control compensation vector and the longitudinal control compensation vector.

[0158] The fifth step involves obtaining the operator and performing smoothing filtering on the synthesized control compensation vector to obtain the target control compensation vector.

[0159] In some alternative implementations, the correction module 604 includes: The transformation unit is used to transform the target control compensation vector into a transformed control compensation vector in the body coordinate system.

[0160] The splitting unit is used to separate the lateral conversion control compensation vector and the longitudinal conversion control compensation vector from the conversion control compensation vector.

[0161] The lateral and longitudinal compensation component calculation unit is used to calculate the lateral compensation component and the longitudinal compensation component based on the lateral conversion control compensation vector and the longitudinal conversion control compensation vector, respectively.

[0162] The roll and pitch correction calculation unit is used to calculate the target roll and pitch corrections based on the lateral and longitudinal compensation components, respectively.

[0163] The original expected roll and pitch angle acquisition unit is used to acquire the original expected roll angle and original expected pitch angle corresponding to the initial trajectory.

[0164] The corrected desired roll angle is obtained by adding the target roll angle correction amount to the original desired roll angle to obtain the corrected desired roll angle.

[0165] The corrected desired pitch angle is obtained by adding the target pitch angle correction amount to the original desired pitch angle to obtain the corrected desired pitch angle.

[0166] The current roll and pitch angle acquisition unit is used to acquire the current roll angle and current pitch angle of the target UAV.

[0167] The roll channel attitude deviation calculation unit is used to calculate the roll channel attitude deviation between the corrected desired roll angle and the current roll angle.

[0168] The pitch channel attitude deviation calculation unit is used to calculate the pitch channel attitude deviation between the corrected desired pitch angle and the current pitch angle.

[0169] The attitude adjustment command generation unit is used to generate attitude adjustment commands based on the attitude deviations of the roll channel and the pitch channel.

[0170] The compensation acceleration determination unit is used to determine the compensation acceleration based on attitude adjustment commands.

[0171] The target trajectory acquisition unit is used to correct the initial trajectory based on the compensation acceleration to obtain the target trajectory corresponding to the target UAV.

[0172] The drone hoisting and transportation sway suppression device provided in this embodiment of the invention can execute the drone hoisting and transportation sway suppression method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0173] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0174] The following is a detailed reference. Figure 7 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes according to a program stored in ROM 702 or a program loaded from memory 708 into RAM 703. RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via bus 704. An input / output (I / O) interface 705 is also connected to bus 704; wherein, ROM is a read-only memory and RAM is a random access memory.

[0175] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0176] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 709, or installed from a memory 708, or installed from a ROM 702. When the computer program is executed by the processor 701, it performs the functions defined in the UAV hoisting and transport sway suppression method of the embodiments of the present invention.

[0177] Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0178] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the UAV hoisting and transport sway suppression method shown in the above embodiments is implemented.

[0179] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0180] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for suppressing swaying during unmanned aerial vehicle (UAV) hoisting and transportation, characterized in that, The method includes: Acquire flight payload images corresponding to the target payload of the target UAV during its flight; Acquire a reference load image corresponding to the target payload when the target UAV is in a stable hovering state; Based on the flight load image and the reference load image, calculate the target control compensation vector corresponding to the target load of the target UAV during flight; Based on the target control compensation vector, the initial trajectory corresponding to the target UAV is corrected to obtain the target trajectory corresponding to the target UAV.

2. The method according to claim 1, characterized in that, The step of calculating the target control compensation vector corresponding to the target payload during the flight of the target UAV based on the flight payload image and the reference payload image includes: Target recognition is performed on the flight payload image to determine the target payload region and the payload outline pixel area corresponding to the target payload in the flight payload image; Based on the target load area, calculate the real-time load swing angle corresponding to the target load during the flight of the target UAV; Target recognition is performed on the reference load image to determine the reference load pixel area and reference physical rope length corresponding to the reference load image; Based on the payload outline pixel area, the reference payload pixel area, and the reference physical rope length, the dynamic rope length corresponding to the target payload during the flight of the target UAV is calculated. The target control compensation vector is calculated based on the real-time swing angle of the load and the dynamic rope length.

3. The method according to claim 2, characterized in that, The step of calculating the real-time load swing angle corresponding to the target load during the flight of the target UAV based on the target load region includes: Obtain the center coordinates of the payload image corresponding to the flight payload image; Based on the target load region, determine the load pixel position coordinates corresponding to the geometric center of the target load; Based on the load pixel position coordinates and the load image center coordinates, a two-dimensional pixel deviation vector is calculated; Based on the lateral deviation vector in the two-dimensional pixel deviation vector, the real-time lateral swing angle corresponding to the target payload of the target UAV during flight is calculated. Based on the longitudinal deviation vector in the two-dimensional pixel deviation vector, the longitudinal real-time swing angle corresponding to the target payload of the target UAV during flight is calculated; The real-time swing angle of the load is calculated based on the real-time lateral swing angle and the real-time longitudinal swing angle.

4. The method according to claim 2, characterized in that, The calculation of the target control compensation vector based on the real-time swing angle of the load and the dynamic rope length includes: Based on the real-time swing angle of the load, calculate the load swing angular velocity corresponding to the target load; An augmented state vector is generated based on the real-time swing angle of the load, the dynamic rope length, and the angular velocity of the load swing. Configure the target gain matrix of the extended state observer based on the dimension of the augmented state vector; The augmented state vector is input into the extended state observer, and a discrete-time recursion is performed based on the target gain matrix to output the total perturbation estimate. Based on the total disturbance estimate, the target control compensation vector is calculated.

5. The method according to claim 4, characterized in that, The step of calculating the target control compensation vector based on the total disturbance estimate includes: The total disturbance estimate is decomposed to obtain the lateral equivalent disturbance component and the longitudinal equivalent disturbance component; Based on the dynamic rope length and gravitational acceleration, the geometric mapping coefficients are calculated; The geometric mapping coefficients are multiplied by the lateral equivalent disturbance component and the longitudinal equivalent disturbance component respectively to obtain the lateral original disturbance acceleration and the longitudinal original disturbance acceleration. The signs of the original lateral perturbation acceleration and the original longitudinal perturbation acceleration are inverted to obtain the corrected lateral perturbation acceleration and the corrected longitudinal perturbation acceleration. Based on the lateral corrected perturbation acceleration and the longitudinal corrected perturbation acceleration, the acceleration correction vector is calculated; The target control compensation vector is obtained based on the acceleration correction vector.

6. The method according to claim 5, characterized in that, The process of obtaining the target control compensation vector based on the acceleration correction vector includes: Obtain the initial control variables corresponding to the target UAV; the initial control variables include initial lateral control variables and initial longitudinal control variables. Obtain the real-time heading angle corresponding to the target UAV; Based on the real-time heading angle, a rotation matrix is ​​constructed from the body coordinate system to the geographic Cartesian coordinate system; Based on the rotation matrix, the initial control quantity is transformed to obtain the transformed control quantity in the geographic Cartesian coordinate system; Based on the lateral correction disturbance acceleration in the acceleration correction vector, the lateral conversion control quantity in the conversion control quantity is corrected to obtain the lateral control compensation vector. Based on the longitudinal correction disturbance acceleration in the acceleration correction vector, the longitudinal conversion control quantity in the conversion control quantity is corrected to obtain the longitudinal control compensation vector; Based on the lateral control compensation vector and the longitudinal control compensation vector, a synthetic control compensation vector is obtained; The synthesized control compensation vector is smoothed and filtered to obtain the target control compensation vector.

7. The method according to claim 1, characterized in that, The step of correcting the initial trajectory of the target UAV based on the target control compensation vector to obtain the target trajectory of the target UAV includes: Transform the target control compensation vector into a transformed control compensation vector in the body coordinate system; The lateral conversion control compensation vector and the longitudinal conversion control compensation vector are separated from the conversion control compensation vector; Based on the lateral conversion control compensation vector and the longitudinal conversion control compensation vector, the lateral compensation component and the longitudinal compensation component are calculated respectively. Based on the lateral compensation component and the longitudinal compensation component, the target roll angle correction and the target pitch angle correction are calculated respectively. Obtain the original expected roll angle and the original expected pitch angle corresponding to the initial trajectory; The target roll angle correction amount is added to the original desired roll angle to obtain the corrected desired roll angle; The target pitch angle correction is added to the original desired pitch angle to obtain the corrected desired pitch angle; Obtain the current roll angle and current pitch angle of the target UAV; Calculate the roll channel attitude deviation between the corrected desired roll angle and the current roll angle; Calculate the pitch channel attitude deviation between the corrected desired pitch angle and the current pitch angle; Based on the roll channel attitude deviation and the pitch channel attitude deviation, an attitude adjustment command is generated; Based on the attitude adjustment command, determine the compensation acceleration; The initial trajectory is corrected based on the compensation acceleration to obtain the target trajectory corresponding to the target UAV.

8. A device for suppressing swaying during unmanned aerial vehicle (UAV) hoisting and transportation, characterized in that, The device includes: The first acquisition module is used to acquire flight payload images corresponding to the target payload of the target UAV during flight. The second acquisition module is used to acquire a reference load image corresponding to the target load when the target UAV is in a stable hovering state; The calculation module is used to calculate the target control compensation vector corresponding to the target load during the flight of the target UAV based on the flight load image and the reference load image; The correction module is used to correct the initial trajectory of the target UAV based on the target control compensation vector, so as to obtain the target trajectory of the target UAV.

9. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the UAV hoisting and transport sway suppression method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the unmanned aerial vehicle hoisting and transport sway suppression method according to any one of claims 1 to 7.

Images