An electric power inspection unmanned aerial vehicle bad weather adaptive flight control system

Abstract

The present application relates to the technical field of unmanned aerial vehicle flight control, and particularly relates to a power inspection unmanned aerial vehicle self-adaptive flight control system in bad weather, which determines a gimbal rotation angle limit polyhedron by three-dimensional boundary projection in a local coordinate system of the machine body; divides a retreat path based on a windward inclination vector of the unmanned aerial vehicle and deduces an expected posture of each test node, constructs a camera line-of-sight sweeping polyhedron containing a local relative line-of-sight, and visualizes a dynamic observation line-of-sight in a future retreat process into a spatial entity in advance; then, the gimbal rotation angle limit polyhedron and the camera line-of-sight sweeping polyhedron are subjected to spatial interference calculation, a direction vector component is extracted to determine an interference direction reference angle, an abstract angle exceeding limit hidden danger is converted into a preposed three-dimensional space collision warning and a future conflict direction is accurately locked; finally, active flight control is performed in advance based on the windward inclination vector of the unmanned aerial vehicle and the interference direction reference angle, so that the flight control reliability in bad weather is higher.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, specifically to an adaptive flight control system for power line inspection UAVs in adverse weather conditions. Background Technology

[0002] In automated inspections of overhead power lines, target towers are often located in mountainous micro-topographical environments prone to gusts of wind. In such environments, when a drone performs tower image acquisition tasks, the fuselage must tilt in the wind to resist airflow disturbances. This requires the gimbal to rotate at a large angle in the opposite direction of the tilt. Current technology typically triggers a hardware over-limit warning when the rotation angle of the motor inside the gimbal reaches its mechanical operating limit and is obstructed, instructing the drone to perform passive flight control by yielding to the wind.

[0003] However, the aforementioned existing technologies do not consider the dynamic predictive interference relationship between the gimbal's mechanical limits, the fuselage's windward attitude, and the target's line of sight in three-dimensional space. They can only react after the gimbal motors are actually obstructed. This results in the system being unable to calculate a targeted interference avoidance direction in advance based on the specific spatial location of the future obstruction of the line of sight. It also prevents the system from avoiding the risk of gimbal exceeding limits through proactive compensation control. Consequently, the adaptive response of the existing technology's passive flight control through triggering limit-exceeding warnings is lagging, and its flight control reliability is poor in adverse weather conditions. Summary of the Invention

[0004] To address the technical problem of poor flight control reliability in adverse weather conditions caused by the lag in adaptive response of existing technologies that rely on triggering over-limit warnings for passive flight control when performing continuous image acquisition tasks on power towers, this invention aims to provide an adaptive flight control system for power line inspection drones in adverse weather conditions. The specific technical solution adopted is as follows:

[0005] The first aspect of this invention provides an adaptive flight control system for a power line inspection drone in adverse weather conditions, comprising:

[0006] The data acquisition module is used to acquire the maximum allowable yaw angle and maximum allowable pitch angle of the UAV-borne optoelectronic gimbal, as well as the current pitch angle and current roll angle of the UAV at the current moment, during the UAV's inspection of the target to be inspected, and to establish a local coordinate system of the UAV with the UAV as the origin.

[0007] The first determining module is used to perform spatial orientation analysis based on the current pitch angle and the current roll angle to determine the UAV's windward tilt vector at the current moment; and to perform three-dimensional boundary projection based on the maximum allowable yaw angle and the maximum allowable pitch angle in the local coordinate system of the aircraft body to determine the gimbal rotation limit polyhedron.

[0008] The second determining module is used to divide the retreat path based on the UAV's windward tilt vector and determine at least two test nodes; to perform expected attitude deduction based on the current pitch angle, current roll angle and position information of each test node and determine the local relative line-of-sight vector corresponding to each test node; and to perform spatial closure processing based on the start and end points of the local relative line-of-sight vectors of all test nodes to construct a camera line-of-sight sweep polyhedron located in the local coordinate system of the aircraft.

[0009] The UAV flight control module is used to perform spatial interference calculations and extract directional vector components based on the gimbal rotation limit polyhedron and the camera line-of-sight sweep polyhedron to determine the interference direction reference angle; and to perform flight control of the power inspection UAV based on the UAV's windward tilt vector and the interference direction reference angle.

[0010] Furthermore, the process of obtaining the UAV's windward tilt vector includes:

[0011] Based on the sine value of the current pitch angle, determine the X-axis component of the UAV's windward tilt vector;

[0012] The Y-axis component of the UAV's windward tilt vector is determined based on the product of the sine of the current roll angle and the cosine of the current pitch angle.

[0013] By zeroing the data, the Z-axis elevation component of the UAV's windward tilt vector is determined.

[0014] A three-dimensional vector is synthesized based on the X-axis direction component, Y-axis direction component, and Z-axis elevation component to generate the UAV's windward tilt vector.

[0015] Furthermore, the process of obtaining the gimbal angle limit polyhedron includes:

[0016] Obtain the reference unit vector directly in front of the UAV in the local coordinate system of the UAV; construct a two-dimensional angle boundary based on the maximum allowable yaw angle and the maximum allowable pitch angle; perform discrete sampling on the two-dimensional angle boundary according to a preset angle sampling interval to determine at least two extreme yaw angles; perform matrix rotation operation on the reference unit vector directly in front based on each extreme yaw angle to determine at least two boundary rays; project the spatial endpoints of each boundary ray based on a preset unit spherical radius constant to determine the discrete spatial endpoints corresponding to each boundary ray; perform spatial convex hull closure processing based on all discrete spatial endpoints to determine the gimbal rotation angle limit polyhedron.

[0017] Furthermore, the process of acquiring the test node includes:

[0018] Obtain the downwind retreat unit vector corresponding to the opposite direction of the UAV's windward tilt vector; obtain the UAV's initial spatial position at the current moment; determine the single-step spatial displacement vector based on the product between the preset single-step sliding length scalar and the downwind retreat unit vector; and determine all test nodes by successively superimposing the position coordinates according to the preset maximum total number of test nodes based on the initial spatial position point and the single-step spatial displacement vector.

[0019] Furthermore, the process of obtaining the local relative line-of-sight vector includes:

[0020] Based on the Euclidean distance between the UAV's initial spatial position and each test node at the current moment, the cumulative slip distance of each test node is determined; the cumulative slip distance is negatively correlated and mapped to determine the attitude decay weight.

[0021] The current pitch angle is weighted based on the attitude attenuation weight to determine the expected pitch angle of each test node; the current roll angle is weighted based on the attitude attenuation weight to determine the expected roll angle of each test node.

[0022] Obtain the target spatial location point of the target to be inspected; use the vector formed by the position coordinates of each test node pointing to the target spatial location point as the global absolute line-of-sight vector corresponding to each test node; construct the inverse pose transformation matrix mapping each test node to the local coordinate system of the aircraft based on the expected pitch angle and expected roll angle of each test node; multiply the global absolute line-of-sight vector and the inverse pose transformation matrix to determine the initial relative line-of-sight vector; multiply the unit vector of the initial relative line-of-sight vector with a preset unit spherical radius constant to determine the local relative line-of-sight vector corresponding to each test node.

[0023] Furthermore, the process of acquiring the camera's line-of-sight sweep of the polyhedron includes:

[0024] In the local coordinate system of the machine body, starting from the origin of the local coordinate system, the local line-of-sight endpoint of each test node is determined based on the local relative line-of-sight vector corresponding to each test node; and a spatial convex hull closure process is performed based on all local line-of-sight endpoints to construct a camera line-of-sight sweep polyhedron.

[0025] Furthermore, the process of obtaining the reference angle of the interference direction includes:

[0026] The gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron are input into the GJK collision interference detection algorithm, and the spatial boundary distance between the gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron and the nearest boundary interference point are output.

[0027] In the local coordinate system of the aircraft, when the spatial boundary spacing is greater than zero, the vector formed by pointing from the origin of the local coordinate system to the nearest boundary interference point is taken as the spatial interference direction vector; when the spatial boundary spacing is equal to zero, the downwind retreat unit vector is taken as the spatial interference direction vector.

[0028] The X-axis projection component and the corresponding Y-axis projection component of the spatial interference direction vector are extracted and subjected to bivariate arctangent operation to determine the reference angle of the interference direction.

[0029] Furthermore, the process of controlling the power line inspection drone's flight based on the drone's windward tilt vector and the interference direction reference angle includes:

[0030] The distance to the target for translational avoidance is determined by multiplying the preset single-step sliding length scalar with the preset maximum total number of measured nodes; the three-dimensional displacement vector is determined by multiplying the downwind retreat unit vector with the distance to the target for translational avoidance; and the coordinates of the target for translational avoidance in the downwind direction are determined by superimposing the coordinates of the initial spatial position point and the three-dimensional displacement vector.

[0031] Based on the product of the interference direction reference angle and the preset yaw feedforward compensation ratio gain, the fuselage yaw compensation rotation angle is determined; the fuselage yaw compensation rotation angle is superimposed with the current yaw orientation of the UAV to determine the nose control orientation command; based on the downwind translation target coordinates and the nose control orientation command, a cooperative compensation flight path command is generated to perform flight control of the power line inspection UAV based on the cooperative compensation flight path command.

[0032] Furthermore, the preset angle sampling interval ranges from 1 degree to 5 degrees.

[0033] Furthermore, the preset maximum total number of measurement nodes is set to 10.

[0034] In a second aspect, the present invention provides a computer device including a memory and a processor. The memory is used to store computer program code, and the processor is used to call and run the computer program code from the memory to execute a system as described in the first aspect or any embodiment of the first aspect of the present invention.

[0035] The present invention has the following beneficial effects:

[0036] This invention determines the gimbal's angle limit polyhedron by projecting three-dimensional boundaries in the local coordinate system of the aircraft, intuitively quantifying the physical limit boundaries of gimbal operation. Based on the UAV's windward tilt vector, it divides the retreat path and extrapolates the expected attitude of each test node, constructing a camera line-of-sight sweep polyhedron containing local relative lines of sight. This visualizes the dynamic observation lines of sight during the future retreat process as spatial entities in advance. Furthermore, it performs spatial interference calculations on the aforementioned gimbal angle limit polyhedron and camera line-of-sight sweep polyhedron, extracting direction vector components to determine the interference direction reference angle. This transforms the abstract angle over-limit risk into a proactive three-dimensional spatial collision warning and accurately locks the future conflict location. Finally, based on the UAV's windward tilt vector and interference direction reference angle, it performs proactive flight control in advance, using the predicted interference location to guide the fuselage for motion compensation. This effectively overcomes the shortcomings of existing technologies that rely on triggering over-limit warnings for passive flight control due to adaptive response lag, resulting in higher reliability of flight control in adverse weather conditions. Attached Figure Description

[0037] Figure 1 This is a structural diagram of an adaptive flight control system for a power line inspection drone in adverse weather conditions, provided as an embodiment of the present invention. Detailed Implementation

[0038] The following description, in conjunction with the accompanying drawings, details a specific scheme for an adaptive flight control system for a power line inspection drone in adverse weather conditions provided by the present invention.

[0039] This invention provides an adaptive flight control system for power line inspection drones in adverse weather conditions. Please refer to [link / reference]. Figure 1 The diagram illustrates a structural diagram of an adaptive flight control system for a power line inspection drone in adverse weather conditions, according to an embodiment of the present invention. The system includes:

[0040] The data acquisition module 101 is used to acquire the maximum allowable yaw angle and maximum allowable pitch angle of the UAV-borne optoelectronic gimbal, as well as the current pitch angle and current roll angle of the UAV at the current moment, during the process of the UAV inspecting the target to be inspected, and to establish a local coordinate system of the UAV with the UAV as the origin.

[0041] During the automated inspection of targets by unmanned aerial vehicles (UAVs), the system first needs to obtain the mechanical and physical limits of the optoelectronic payload hardware itself. Specifically, an optoelectronic gimbal is fixedly mounted under the fuselage of the UAV, and the drive motor inside the gimbal is limited by the physical travel of the factory-designed mechanical structure. During the takeoff preparation phase of each inspection mission, the onboard computer sends parameter query commands to the main control board of the optoelectronic gimbal through the internal communication bus interface, or directly reads the firmware of the hardware configuration inside the optoelectronic gimbal, thereby accurately extracting the maximum allowable yaw angle of the gimbal motor in the horizontal rotation direction and the maximum allowable pitch angle in the vertical rotation direction. These extracted static allowable angles are stored in the running memory as the absolute data benchmark for subsequently constructing the physical operating limits of the gimbal.

[0042] During the dynamic flight of the drone as it approaches the target tower and encounters gusts of wind, the system needs to perceive the drone's attitude response and spatial position in real time. An inertial measurement unit (IMU) is installed at the flight control motherboard, located at the drone's center of gravity. This IMU uses a built-in microelectromechanical gyroscope and accelerometer to monitor the drone's angular velocity and acceleration at high frequency. Based on this, the system analyzes and obtains the drone's current yaw, pitch, and roll angles in real time. Simultaneously, a global positioning system (GPS) receiver mounted on the top of the drone continuously captures satellite signals to calculate real-time latitude, longitude, and altitude data. Using the drone's current instantaneous spatial position as the absolute origin, the system uses a coordinate projection transformation algorithm to uniformly convert the surrounding global latitude and longitude coordinates into Cartesian spatial coordinates measured in meters. This establishes a local coordinate system with the drone as the origin, providing a unified scale and spatial reference for subsequent three-dimensional spatial vector calculations and envelope interferometry measurements.

[0043] The first determining module 102 is used to perform spatial orientation analysis based on the current pitch angle and the current roll angle to determine the UAV's windward tilt vector at the current moment; and to perform three-dimensional boundary projection in the local coordinate system of the aircraft based on the maximum allowable yaw angle and the maximum allowable pitch angle to determine the gimbal rotation limit polyhedron.

[0044] Considering that when a drone maintains hovering operations in a gusty environment, in order to resist the continuous thrust of the external airflow, the fuselage will inevitably tilt in accordance with the wind direction to resist the wind. This attitude change directly reflects the current blowing direction of the airflow and the intensity of the environmental airflow disturbance. Therefore, the system needs to convert the real-time attitude angle values ​​into three-dimensional physical quantities with clear spatial orientation. Therefore, this embodiment of the invention performs spatial orientation analysis based on the current pitch angle and the current roll angle to determine the drone's windward tilt vector at the current moment. By extracting the drone's windward tilt vector, the system can accurately anchor the actual force orientation of the fuselage resisting the gusts, thereby providing a reliable directional guidance benchmark for planning a safe horizontal retreat path in accordance with the airflow.

[0045] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the UAV's windward tilt vector includes:

[0046] Based on the sine of the current pitch angle, the X-axis component of the UAV's tilt vector is determined; based on the product of the sine of the current roll angle and the cosine of the current pitch angle, the Y-axis component of the UAV's tilt vector is determined; through zeroing, the Z-axis elevation component of the UAV's tilt vector is determined; based on the X-axis, Y-axis, and Z-axis elevation components, a three-dimensional vector is synthesized to generate the UAV's tilt vector.

[0047] Considering that the complex tilt state of the aircraft in three-dimensional space is formed by the superposition of forward and backward pitch and left and right roll, the orthogonal decomposition principle of spatial geometry can be used to accurately separate the attitude tendency of the UAV on different spatial coordinate axes through trigonometric function calculations. Furthermore, in mountainous micro-topography environments, strong vertical rise and fall disturbances are often present during inspections. Introducing attitude fluctuations in the elevation direction into vector calculations could easily lead to erroneous rapid climbs or dives during the retreat process, posing a risk of crash. Therefore, the system extracts the sine of the current pitch angle and the product of the sine of the current roll angle and the cosine of the current pitch angle, accurately calculating the directional components in two dimensions on the horizontal plane, and forcibly zeroing out the vertical elevation component, thus synthesizing a complete three-dimensional vector. This design logic combining decomposition and zeroing accurately quantifies the precise direction of the UAV's resistance to gusts in a purely horizontal plane, effectively eliminating the fatal interference of vertical rise and fall airflow on the horizontal retreat trajectory planning, and ensuring that the generated UAV's windward tilt vector always points to the safe horizontal slip limit.

[0048] Since the maximum allowable yaw angle and maximum allowable pitch angle of the opto-gimbal are merely isolated one-dimensional numerical limits, and subsequent spatial interferometry detection algorithms cannot directly calculate the three-dimensional collision distance from the abstract angle values, this invention, in order to concretize the abstract angle constraints into a physical reference object that can be used for three-dimensional geometric calculation, projects a three-dimensional boundary based on the maximum allowable yaw angle and maximum allowable pitch angle in the local coordinate system of the aircraft to determine the gimbal's rotation angle limit polyhedron. By projecting and enclosing the purely hardware numerical limits into a three-dimensional geometric polyhedron within the aircraft's space, the system can intuitively quantify the absolute physical dead zone boundary of the opto-gimbal's allowed operation in three-dimensional space, thereby providing an accurate three-dimensional physical reference standard for subsequent precise calculation of whether the camera's line of sight touches the mechanical limits.

[0049] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the gimbal angle limit polyhedron includes:

[0050] Obtain the reference unit vector directly in front of the UAV in the local coordinate system, i.e., the unit vector pointing directly in front of the UAV's nose; construct a two-dimensional angle boundary based on the maximum allowable yaw angle and the maximum allowable pitch angle; perform discrete sampling on the two-dimensional angle boundary according to a preset angle sampling interval to determine at least two extreme yaw angles; perform matrix rotation operation on the reference unit vector directly in front based on each extreme yaw angle to determine at least two boundary rays; project the spatial endpoints of each boundary ray based on a preset unit spherical radius constant to determine the discrete spatial endpoints corresponding to each boundary ray; perform spatial convex hull closure processing based on all discrete spatial endpoints to determine the gimbal rotation angle limit polyhedron.

[0051] In this embodiment of the invention, the specific process of performing matrix rotation operations on the reference unit vector directly in front based on each extreme deflection angle to determine at least two boundary rays includes: for any extreme deflection angle, extracting the yaw angle component and pitch angle component corresponding to the extreme deflection angle; constructing a yaw rotation matrix around the vertical axis of the aircraft based on the yaw angle component; constructing a pitch rotation matrix around the lateral axis of the aircraft based on the pitch angle component; performing matrix multiplication of the yaw rotation matrix and the pitch rotation matrix to determine the comprehensive rotation matrix corresponding to the extreme deflection angle; performing matrix multiplication of the comprehensive rotation matrix and the reference unit vector directly in front to determine the deflection direction vector corresponding to the extreme deflection angle; and generating the boundary ray corresponding to the extreme deflection angle with the origin of the local coordinate system of the aircraft as the starting point and the direction of the deflection direction vector as the extension direction.

[0052] In one specific implementation of this invention, the preset angle sampling interval is set to a range of 1 to 5 degrees, which can be adjusted according to the specific implementation environment. In this embodiment, the preset angle sampling interval is set to 1 degree. The preset angle sampling interval is used to characterize the spatial fineness when performing three-dimensional discretization modeling of the physical operating boundary of the gimbal. When the computing power of the UAV's onboard computer is limited, or when encountering high-frequency gust disturbances that cause the control system to prioritize the real-time response of the collision calculation algorithm, the value of the preset angle sampling interval should be larger (e.g., 5 degrees) to reduce the number of discrete spatial endpoints generated, thereby reducing the computing power overhead of spatial convex hull closure and subsequent interference calculation. Conversely, when the computing power of the onboard computer is sufficient, or when the system has extremely high requirements for the triggering accuracy of the gimbal collision warning, the value of the preset angle sampling interval should be smaller (e.g., 1 degree in this embodiment) to generate high-density discrete spatial endpoints, thereby restoring the real smooth spherical physical boundary of the optoelectronic gimbal as much as possible and avoiding interference misjudgment caused by the roughness of the polyhedral edges.

[0053] In one specific implementation of this invention, the preset unit spherical radius constant is set to a range of 0.5 meters to 2.0 meters, which can be adjusted according to the specific implementation environment. In this embodiment, the preset unit spherical radius constant is set to 1.0 meter. The preset unit spherical radius constant is used to characterize the virtual space projection depth referenced when the abstract angle limit of the gimbal is transformed into a three-dimensional geometric entity in the local coordinate system of the drone. When the drone body is large, the physical size of the under-mounted optoelectronic gimbal is long, or the control system needs to enlarge the geometric size of the warning boundary to improve the overall obstruction safety margin when performing spatial interference calculations, the value of the preset unit spherical radius constant should be large (e.g., 2.0 meters). Conversely, when the drone is a micro-small and lightweight model, the optoelectronic gimbal structure is highly compact, or the system needs to allow the camera to work in a state closer to the mechanical blind spot when performing complex trajectory tracking, the value of the preset unit spherical radius constant should be small (e.g., 0.5 meters).

[0054] Considering that the actual operating space of an electro-optical gimbal is a three-dimensional spherical region with continuous and smooth transitions, simply relying on the extreme points formed by the maximum allowable yaw and pitch angles for simple envelope construction would generate pyramidal geometries with numerous blind spots, severely deviating from the actual mechanical motion trajectory of the electro-optical gimbal. To eliminate spatial calculation errors caused by coarse modeling at extreme points, the system needs to perform meticulous discretization within the angular boundary range. Therefore, based on the reference unit vector directly in front of the UAV, the system performs dense discrete sampling at preset angle sampling intervals on the two-dimensional angular boundary formed by the maximum allowable yaw and pitch angles to generate dense limit yaw angles, and uses matrix rotation operations to generate multiple boundary rays pointing to the limit azimuth. Furthermore, considering that the boundary rays lack truncation depth coordinate support in three-dimensional space, the system introduces a fixed preset unit spherical radius constant to truncate the boundary rays into discrete spatial endpoints with specific depths. Finally, spatial convex hull closure is used to enclose all discrete spatial endpoints into a three-dimensional envelope with a defined volume. The design logic of combining dense discrete sampling with spherical endpoint projection and closure accurately restores the real smooth operating boundary of the optoelectronic gimbal under extreme deflection state, effectively eliminating spatial dead angle loopholes caused by extreme point modeling, and ensuring that the generated gimbal rotation limit polyhedron can accurately represent the three-dimensional spatial morphology of the physical limit operation of the optoelectronic gimbal.

[0055] The second determining module 103 is used to divide the retreat path based on the UAV's windward tilt vector and determine at least two test nodes; to perform expected attitude deduction based on the current pitch angle, current roll angle and position information of each test node and determine the local relative line-of-sight vector corresponding to each test node; and to perform spatial closure processing based on the start and end points of the local relative line-of-sight vectors of all test nodes to construct a camera line-of-sight sweep polyhedron located in the local coordinate system of the aircraft.

[0056] While the previously obtained UAV tilt vector quantifies the UAV's attitude orientation against gusts at the current moment, the spatial direction component alone cannot deduce the dynamic spatial position change of the UAV during the downwind translation and force relief process. Therefore, this embodiment of the invention considers that the UAV's backward retreat in a gusty environment is a dynamic process of continuous displacement in three-dimensional space, and it is necessary to predict the risk of interference with the tracking line of sight of the electro-optical gimbal in advance on the continuous sliding path. Therefore, the retreat path is further divided based on the UAV tilt vector, and at least two test nodes are determined. By discretizing the continuous retreat action into a series of specific spatial coordinate points, the system can construct a future path model for attitude deduction along the wind direction, thereby providing a discretized physical position basis for the subsequent establishment of a three-dimensional entity envelope.

[0057] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the test node includes:

[0058] Obtain the downwind retreat unit vector corresponding to the opposite direction of the UAV's windward tilt vector, i.e., the unit vector corresponding to the opposite direction of the UAV's windward tilt vector; obtain the UAV's initial spatial position point at the current moment; determine the single-step spatial displacement vector based on the product between the preset single-step sliding length scalar and the downwind retreat unit vector; and determine all test nodes by successively superimposing the position coordinates according to the preset maximum total number of test nodes based on the initial spatial position point and the single-step spatial displacement vector.

[0059] Considering that the drone tilts towards the wind when resisting wind, a reasonable retreat strategy is to move backward in the direction of the wind to reduce air resistance caused by strong winds and lower the output power of the rotor motors. Simultaneously, the onboard computer cannot directly process an infinitely continuous physical displacement trajectory; therefore, the continuous retreat trajectory needs to be converted into a finite number of discrete sampled coordinates to control computational overhead. Thus, the system reverses and normalizes the length of the drone's tilt vector towards the wind to obtain a downwind retreat unit vector indicating the retreat direction. Subsequently, the system introduces a fixed, preset single-step sliding length scalar as the spatial step size for discretization sampling. This preset single-step sliding length scalar is multiplied by the downwind retreat unit vector, transforming the direction vector into a single-step spatial displacement vector with a distance value. Next, using the drone's current initial spatial position as the spatial starting point, the system performs a successive accumulation of spatial coordinates along the direction indicated by the downwind retreat unit vector, using a preset maximum total number of measurement nodes as an iterative constraint. This design logic, based on the superposition of wind direction reversal and fixed step-size iteration, quantifies the spatial position of the UAV at each stage of its tailwind unloading process. This allows the previously unknown continuous sliding trajectory to be explicitly resolved into multiple ordered test nodes. The path partitioning method reduces the complexity of dynamic motion prediction into a discrete coordinate sequence that is easy to compute, effectively controlling the computational power consumption of the spatial interferometric simulation process.

[0060] In one specific implementation of this invention, the preset single-step sliding length scalar is set to a range of 0.1 meters to 1.0 meters, which can be adjusted according to the specific implementation environment. In this embodiment, the preset single-step sliding length scalar is set to 0.5 meters. The preset single-step sliding length scalar is used to characterize the single step span when the continuous retreat path is spatially discretized and sampled. When the UAV encounters strong gusts that result in a long expected translational retreat space span, or when the onboard computer needs to reduce the computational load on the path nodes, the value of the preset single-step sliding length scalar should be larger (e.g., 1.0 meters). Conversely, when the obstacles in the inspection environment are densely distributed, or when the control system needs to perform high-frequency and dense spatial verification of the camera's future line-of-sight interference, the value of the preset single-step sliding length scalar should be smaller (e.g., 0.1 meters).

[0061] In one specific implementation of this invention, the preset maximum total number of calculation nodes is set to a range of 5 to 20, which can be adjusted according to the specific implementation environment. In this embodiment, the preset maximum total number of calculation nodes is set to 10. The preset maximum total number of calculation nodes is used to characterize the scale of the spatial node sequence for the system's forward-looking simulation of the UAV's downwind translation process. When the UAV needs to reserve a longer safe retreat depth in a strong airflow environment, or when the system has high requirements for the spatial warning span of the gimbal obstruction risk, the preset maximum total number of calculation nodes should be larger (e.g., 20). Conversely, when the UAV's onboard processor has limited computing resources, or when the encountered airflow disturbance is weak and the expected retreat avoidance range is small, the preset maximum total number of calculation nodes should be smaller (e.g., 5).

[0062] Considering that during the downwind retreat of the UAV, as the speed of the fuselage relative to the ambient airflow decreases, the tilt attitude required by the fuselage to resist the airflow will gradually return to a smooth transition state. Furthermore, throughout the retreat process, the electro-optical gimbal needs to continuously aim at the stationary target to be inspected. Therefore, simply knowing the change in spatial position is insufficient to determine the mechanical obstruction inside the gimbal. The system needs to combine the position change with the dynamic recovery of the fuselage attitude. Thus, this embodiment of the invention performs expected attitude deduction based on the current pitch angle, current roll angle, and position information of each test node to determine the local relative line-of-sight vector corresponding to each test node. By inversely mapping the observation line of sight in the global space to the local space of the fuselage that tilts with each test node, the system can obtain the relative deflection state of the camera inside the fuselage in order to align with the target during the future retreat phase. This lays a rigorous mathematical foundation for subsequently transforming the abstract tracking line of sight into a three-dimensional entity in the same reference system as the gimbal's corner limit polyhedron.

[0063] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the local relative line-of-sight vector includes:

[0064] Based on the Euclidean distance between the initial spatial position of the UAV at the current moment and each test node, the cumulative slip distance of each test node is determined; the cumulative slip distance is negatively correlated and mapped to determine the attitude attenuation weight. Specifically, in this embodiment of the invention, a preset maximum retreat reference distance is extracted, the ratio between the cumulative slip distance and the preset maximum retreat reference distance is calculated, and the negative sign of the ratio is used as the exponent of an exponential function with the natural constant as the base. The exponentiation is then performed to obtain the attitude attenuation weight with a value range between 0 and 1. In one specific implementation of this invention, the preset maximum retreat reference distance is set to a range of 3 to 8 meters, which can be adjusted according to the specific implementation environment. In this embodiment, the preset maximum retreat reference distance is set to 5 meters. The preset maximum retreat reference distance is used to characterize the degree of smoothness of the spatial transition as the aircraft gradually decays and returns to a horizontal state during the tailwind retreat translation process. When the aircraft's frontal area is large or it encounters a continuous high-intensity stable airflow, the preset maximum retreat reference distance should be larger (e.g., 8 meters). Conversely, when the aerodynamic drag coefficient of the UAV is small or the wind field in the inspection environment is severely attenuated, the preset maximum retreat reference distance should be smaller (e.g., 3 meters).

[0065] The expected pitch angle of each test node is determined by weighting the current pitch angle based on the attitude attenuation weight; the expected roll angle of each test node is determined by weighting the current roll angle based on the attitude attenuation weight; in this embodiment of the invention, the expected pitch angle of each test node is determined by the product of the attitude attenuation weight and the current pitch angle, and the expected roll angle of each test node is determined by the product of the attitude attenuation weight and the current roll angle.

[0066] The process involves: acquiring the target spatial location of the target to be inspected; using the vector formed by the position coordinates of each test node pointing to the target spatial location as the global absolute line-of-sight vector for each test node; constructing an inverse pose transformation matrix mapping each test node to the local coordinate system based on the expected pitch and roll angles of each test node; specifically, extracting the current yaw angle, expected pitch angle, and expected roll angle of each test node, and generating the corresponding positive attitude rotation matrix using Euler angle rotation rules; performing pose inversion processing on the positive attitude rotation matrix using quaternion inversion operations to generate the inverse pose transformation matrix for each test node; multiplying the global absolute line-of-sight vector and the inverse pose transformation matrix to determine the initial relative line-of-sight vector; and multiplying the unit vector of the initial relative line-of-sight vector with a preset unit spherical radius constant to determine the local relative line-of-sight vector for each test node. It should be noted that the method of generating the positive attitude rotation matrix through Euler angle rotation rules and the process of generating the inverse pose transformation matrix using quaternion inversion are techniques well known to those skilled in the art, and will not be further limited or elaborated here.

[0067] Considering the nonlinear decay of the fuselage attitude recovery process during tailwind shearing, the system introduces exponential operations to calculate the attitude decay weights. The cumulative slip distance is compared with a fixed, preset maximum yield reference distance using a division ratio and then exponentially mapped to quantify the weakening effect of the yield on air resistance. Furthermore, since the UAV's tilt resistance in three-dimensional space is a combination of pitch motion in the forward / backward dimension and roll motion in the left / right dimension, when the overall drag coefficient decreases due to tailwind shearing, the resistance forces on the fuselage in these two dimensions will also decrease to the same extent. Therefore, the calculated attitude decay weights are directly multiplied and weighted by the current pitch angle and current roll angle respectively. Using the same rigorously mapped decay ratio, the expected pitch angle and expected roll angle of the fuselage at various future nodes are simultaneously derived, achieving a reliable prediction of the smooth and gradual decrease of the fuselage's three-dimensional tilt state at different slip nodes.

[0068] Since the target to be inspected remains stationary in the objective environment, the global absolute line-of-sight vectors pointing from each test node to the target's spatial location only reflect the pure geometric connections in absolute space, while the mechanical collision boundary of the electro-optical gimbal is fixed to the UAV's fuselage. If collision detection is performed directly in the global coordinate system, the detection model will be detached from the tilt state of the fuselage itself, losing the practical significance of hardware obstruction warning. Therefore, the system uses the derived expected pitch and roll angles to establish an inverse pose transformation matrix representing the spatial reversal relationship. Using the inverse pose transformation matrix to rotate and transform the global absolute line-of-sight vector essentially forces the absolute connections of the external environment into the internal view of the UAV's tilting and swaying. This reverse mapping logic from global to local eliminates the spatial interference of fuselage attitude changes on the camera's observation direction, allowing the extracted local relative line-of-sight vectors to truly reflect the relative deflection angle that the internal drive motor of the electro-optical gimbal will face, ensuring the logical rigor and reliability of the subsequent three-dimensional spatial collision warning at the physical level.

[0069] Considering that the length of the global absolute line-of-sight vector depends on the actual physical distance between the UAV and the target to be inspected, and the gimbal corner limit polyhedron is constructed based on a preset unit spherical radius constant, directly using a large-scale line-of-sight vector to construct the sweeping polyhedron would cause the line-of-sight geometry to directly penetrate the limit boundary volume during collision interference calculations, leading to calculation deadlock or algorithm failure. Therefore, the system extracts the unit vector of the initial relative line-of-sight vector and multiplies it by the preset unit spherical radius constant for distance truncation. This logic of scale alignment and distance truncation proportionally shrinks the long-distance tracking line of sight onto a virtual sphere identical to the limit boundary, completely eliminating the geometric penetration risk caused by significant scale differences and ensuring the scale equivalence of the two polyhedra in interference comparison calculations.

[0070] Furthermore, considering that the optoelectronic gimbal will undergo continuous viewpoint deflection when following the drone's retreat, and that isolated line-of-sight vectors cannot reflect the camera's three-dimensional spatial occupancy during the entire retreat process, and cannot be directly used as input for subsequent three-dimensional collision interference detection algorithms, in order to visualize the camera's line-of-sight changes during dynamic tracking as a three-dimensional entity that can be geometrically measured, this embodiment of the invention performs spatial closure processing based on the start and end points of the local relative line-of-sight vectors of all test nodes, constructing a camera line-of-sight sweeping polyhedron located in the local coordinate system of the aircraft; by enclosing the discrete line-of-sight endpoints together with the origin of the spatial coordinates to form a closed volume, the system can intuitively restore the complete spatial area swept by the camera's optical axis during continuous retreat tracking, thereby providing a reliable three-dimensional entity reference standard for subsequent calculation of the collision interference between the dynamic line of sight and the mechanical physical dead zone.

[0071] Preferably, in some possible implementations of the embodiments of the present invention, the process of acquiring the polyhedron by camera line of sight includes:

[0072] In the local coordinate system of the machine body, starting from the origin of the local coordinate system, the local line-of-sight endpoint of each test node is determined based on the local relative line-of-sight vector corresponding to each test node; and a spatial convex hull closure is performed based on all local line-of-sight endpoints to construct the camera line-of-sight sweep polyhedron.

[0073] Considering that the local coordinate system of the aircraft is established based on the current location of the UAV, its spatial origin physically corresponds to the mounting hinge center of the electro-optical gimbal under the UAV's fuselage, i.e., the launch starting point of the camera's line of sight. If only the line of sight directions of each test node are analyzed spatially, it is difficult to construct a geometric entity with physical volume. Therefore, in this embodiment of the invention, the origin of the local coordinate system is used as the physical starting point, and the local relative line of sight vector corresponding to each test node is extended to extract the endpoint of each local line of sight.

[0074] Furthermore, if only the local line-of-sight endpoints at the far end are used for closure, the generated geometry will appear as a thin, folded surface suspended in three-dimensional space. This violates the real physical spatial law that the camera's line of sight is continuously emitted from the lens origin, and will also cause the subsequent three-dimensional geometric interferometry detection algorithm to directly penetrate and fail. To eliminate this hidden danger, this embodiment of the invention merges the origin of the local coordinate system of the machine body, representing the starting point of the optical axis emission, with all the local line-of-sight endpoints representing the end of the optical axis sweep, and uses them together as the input point set for spatial convex hull closure processing. The spatial convex hull closure algorithm is used to wrap these discrete coordinate points, forming a radial geometric entity with the gimbal mounting center as the endpoint. This design logic of merging and closing the rotation starting point and the sweep end effectively fills the spatial gap in the line-of-sight transition area between discrete test nodes, so that the constructed camera line-of-sight sweep polyhedron can realistically restore the total volume of three-dimensional space occupied by the camera to track the target during the entire retreat process, ensuring the rigor of the collision interferometry calculation model at the geometric and physical levels.

[0075] The UAV flight control module 104 is used to perform spatial interference calculations and extract directional vector components based on the gimbal rotation limit polyhedron and the camera line-of-sight sweep polyhedron to determine the interference direction reference angle; and to perform flight control of the power inspection UAV based on the UAV's windward tilt vector and the interference direction reference angle.

[0076] The gimbal angle limit polyhedron represents the maximum three-dimensional physical space boundary that the opto-gimbal is allowed to operate in terms of hardware structure. The camera line-of-sight sweep polyhedron represents the total three-dimensional space volume that the opto-gimbal needs to occupy to continuously track the target during subsequent retreat. In order to determine in advance whether the camera will touch the internal mechanical dead zone of the gimbal during dynamic target tracking, and considering that collision prediction in three-dimensional space requires transforming the abstract angle limit risk into an intuitive geometric spatial overlap calculation, and that the three-dimensional properties of spatial interference need to be reduced in dimensionality to be transformed into a control basis that can guide the fuselage to yaw in the horizontal plane, this embodiment of the invention further performs spatial interference calculation and extracts direction vector components based on the gimbal angle limit polyhedron and the camera line-of-sight sweep polyhedron to determine the interference direction reference angle. By performing geometric interference calculation and direction dimensionality reduction extraction on two three-dimensional polyhedra representing different physical meanings, the system can transform spatial collision warning into a specific horizontal plane yaw angle reference, thereby providing a clear quantitative control direction for subsequent active control of fuselage rotation to preemptively share the gimbal load.

[0077] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the interference direction reference angle includes:

[0078] The gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron are input into the GJK collision interferometry detection algorithm, which outputs the spatial boundary distance between the gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron, as well as the nearest boundary interference point. Specifically, the processing of the GJK collision interferometry detection algorithm includes: extracting the coordinate sets of all vertices on the outer surface of the gimbal corner limit polyhedron and the coordinate sets of all vertices on the outer surface of the camera line-of-sight sweep polyhedron; performing a difference operation on the two sets of vertex coordinates to construct the corresponding Minkowski difference graph; generating an initial simplex in the three-dimensional space enclosing the difference graph, and driving the initial simplex to iteratively approach the origin of the coordinate system; when the initial simplex moves to the limit approximation position and the iteration converges, extracting the Euclidean distance on the simplex surface closest to the origin of the coordinate system as the spatial boundary distance, and mapping the corresponding nearest point on the simplex surface back to the surface of the original geometry as the nearest boundary interference point. It should be noted that the GJK collision interferometry detection algorithm is a well-known technique in the art, and its specific process and function will not be further limited or elaborated here.

[0079] Considering that both the gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron are complex 3D geometric figures composed of discrete spatial vertices, using conventional triangular facet traversal intersection calculation methods would result in enormous airborne computing power consumption. Therefore, the system introduces the GJK collision interferometry detection algorithm, leveraging its low computational overhead characteristic for rapid ranging of convex polyhedra. Through difference set construction and iterative approximation of the figure, it outputs the spatial boundary distance reflecting the geometric proximity of the two polyhedra, as well as the nearest boundary interference point indicating the collision location. This calculation method reduces the computational load on the airborne computer and ensures the real-time performance of over-limit collision avoidance warnings.

[0080] In the local coordinate system of the machine body, when the spatial boundary spacing is greater than zero, the vector formed by pointing from the origin of the local coordinate system to the nearest boundary interference point is taken as the spatial interference direction vector; when the spatial boundary spacing is equal to zero, the downwind retreat unit vector is taken as the spatial interference direction vector; the X-axis projection component and the corresponding Y-axis projection component of the spatial interference direction vector are extracted and subjected to bivariate arctangent operation to determine the reference angle of the interference direction.

[0081] Considering that when two polyhedra have not yet interfered, the line connecting the origin to the nearest boundary interference point can indicate the spatial orientation most likely to collide in the future; however, when the two have already overlapped and interfered, making the spatial boundary distance zero, directly extracting the boundary interference line will result in an invalid vector with zero length, thus causing abnormal crashes in the subsequent angle calculation process. Therefore, this embodiment of the invention directly calls the downwind yield unit vector that follows the safe wind direction as a collision avoidance alternative orientation when intersecting interference occurs, ensuring the stable output of the spatial interference direction vector under various extreme interference states. Furthermore, since the yaw control command of the flight controller is executed on a two-dimensional horizontal plane, a three-dimensional vector containing elevation information cannot directly drive the fuselage to yaw, and traditional single-variable arctangent mathematical operations have the risk of numerical overflow with a zero denominator when dealing with coordinate axis crossings. Therefore, this embodiment of the invention removes vertical elevation interference, extracts only the X-axis projection component and the corresponding Y-axis projection component on the horizontal plane, and calls the bivariate arctangent operation for mathematical transformation. This design logic, which combines dimensionality reduction mapping with bivariate operations, not only eliminates the risk of computational overflow, but also accurately identifies the specific spatial quadrant where the geometric interference is located, thereby unambiguously transforming the abstract three-dimensional interference orientation into the interference direction reference angle that guides the horizontal rotation of the UAV fuselage.

[0082] Finally, since the obtained interference direction reference angle and tailwind retreat unit vector have clearly indicated the appropriate movement direction for the fuselage to avoid over-limit risks and air resistance, and since directly executing maneuver commands in a strong gust environment requires consideration of the physical environment's collision safety and the underlying control logic of the flight controller, this embodiment of the invention uses the UAV's windward tilt vector and interference direction reference angle for power inspection UAV flight control. Through the coordinated action of translation retreat unloading and active fuselage yaw, the hidden dangers of full-load rotor operation and gimbal mechanical obstruction are effectively eliminated, ensuring the safe closed loop of the image tracking mission.

[0083] Preferably, in some possible implementations of the embodiments of the present invention, the process of flight control of the power inspection drone based on the drone's windward tilt vector and interference direction reference angle includes:

[0084] The distance to the target for translational avoidance is determined by multiplying the preset single-step sliding length scalar with the preset maximum total number of calculated nodes. A three-dimensional displacement vector is determined by multiplying the downwind retreat unit vector with the distance to the target for translational avoidance. The coordinates of the target for translational avoidance in the downwind direction are determined by superimposing the initial spatial position point with the three-dimensional displacement vector. Specifically, in this embodiment, the relative distance to the nearest obstacle along the downwind retreat unit vector is obtained from the airborne rear-view obstacle avoidance radar sensor. The available retreat margin is calculated by subtracting the flight safety radius constant from this relative distance, and is forcibly set to zero when the available retreat margin is less than zero. It is determined whether the distance to the target for translational avoidance is greater than the available retreat margin. If it is, the distance to the target for translational avoidance is forcibly truncated to the available retreat margin. Subsequently, the three-dimensional displacement vector is determined by multiplying the truncated and corrected distance to the target for translational avoidance with the downwind retreat unit vector, and is superimposed with the initial spatial position point to generate the coordinates of the target for translational avoidance in the downwind direction.

[0085] In one specific implementation of this invention, the flight safety radius constant is set to a range of 1 to 3 meters, which can be adjusted according to the specific implementation environment. In this embodiment, the flight safety radius constant is set to 2 meters. The flight safety radius constant is used to characterize the minimum physical isolation distance that the UAV must maintain from the obstacle behind it when it is retreating to avoid it. When the physical size of the UAV is large, the wind direction and turbulence of the inspection environment are complex, or the obstacle behind it is a high-risk facility such as a high-voltage power transmission cable, the value of the flight safety radius constant should be larger (for example, 3 meters) to reserve a safe buffer space to cope with inertial slip and sudden changes in wind force. Conversely, when the UAV is lightweight, the ambient airflow is relatively stable, or the obstacle behind it is a low-risk obstacle such as a gentle slope, the value of the flight safety radius constant should be smaller (for example, 1 meter).

[0086] Considering the serious safety hazard of blindly commanding a drone to retreat based solely on wind direction in a mountainous micro-topography environment, potentially leading to collisions with mountainsides or power lines, a physical space safety boundary verification mechanism must be introduced before synthesizing the final target coordinates. Therefore, this embodiment of the invention utilizes an airborne rear-view obstacle avoidance radar sensor to detect the real environment. By calculating the difference between the relative distance and the constant flight safety radius, the absolute physical limit of the actual allowable backward sliding of the drone is quantified, i.e., the usable retreat margin. Furthermore, by judging the relationship between the theoretically calculated translational avoidance target distance and this physical limit, and by forcibly truncating the margin to a usable retreat margin when it exceeds the limit, and by providing zero-value fallback protection, the spatial overlap risk caused by blind calculation in the global coordinate domain can be effectively eliminated. This design logic of one-dimensional scalar comparison and forced truncation correction effectively overcomes the risk of negative retreat and collisions that are easily caused by traditional retreat strategies, thereby ensuring that the final generated downwind translational target coordinates are always within a safe physical isolation range.

[0087] The fuselage yaw compensation rotation angle is determined by multiplying the interference direction reference angle with the preset yaw feedforward compensation proportional gain. This yaw compensation rotation angle is then superimposed with the UAV's current yaw orientation to determine the nose control orientation command. Based on the downwind translation target coordinates and the nose control orientation command, a cooperative compensation flight path command is generated for flight control of the power line inspection UAV. In this embodiment, the UAV's current yaw orientation is the absolute heading angle of the UAV's nose at the current moment, which can be determined through real-time monitoring data from the onboard inertial measurement unit and electronic compass.

[0088] In one specific implementation of this invention, the preset yaw feedforward compensation ratio gain is set to a value range of 0.5 to 1.0, which can be adjusted according to the specific implementation environment. In this embodiment, the preset yaw feedforward compensation ratio gain is set to 0.8. The preset yaw feedforward compensation ratio gain is used to characterize the expected sharing ratio of the active yaw rotation action of the fuselage to the future interference deflection of the opto-gimbal. When the remaining redundant stroke of the motor inside the opto-gimbal is long, or the rotational inertia of the UAV itself is large, resulting in excessive energy consumption due to frequent yaw, the value of the preset yaw feedforward compensation ratio gain should be small (e.g., 0.5) to reduce the intervention intensity of the fuselage. Conversely, when the gimbal motor is easily overloaded and obstructed, or when the gust wind speed causes the predicted collision interference to deepen sharply, the value of the preset yaw feedforward compensation ratio gain should be large (e.g., 1.0) to instruct the fuselage to bear the rotation demand of the camera to the greatest extent.

[0089] Considering that the calculated interference direction reference angle reflects the pure geometric mapping characteristics of the three-dimensional collision orientation, directly issuing it as a yaw control quantity could easily lead to overshoot or undershoot in the fuselage rotation. Therefore, the system introduces a preset yaw feedforward compensation proportional gain to linearly scale the geometric reference angle, transforming it into a fuselage yaw compensation rotation angle with actual flight control execution dimensions. Furthermore, since the fuselage yaw compensation rotation angle is a rotational offset relative to the current fuselage state, and the underlying flight controller typically requires a clear absolute heading angle input when performing waypoint navigation, this embodiment of the invention algebraically superimposes the calculated relative offset with the UAV's current yaw orientation. This design logic, based on the algebraic superposition of relative offset and current absolute orientation, seamlessly converts the abstract interference avoidance angle into an absolute nose control orientation command that the flight controller can directly recognize and execute, ensuring the accurate execution of the fuselage coordinated yaw action in the underlying flight control logic.

[0090] In one specific implementation of this invention, after controlling the power inspection drone's flight based on the drone's windward tilt vector and interference direction reference angle, the method further includes: obtaining the initial dead zone radius of the flight controller; adding the target avoidance distance to the initial dead zone radius to generate a target position tolerance range and writing it into the flight controller; obtaining the drone's current global positioning coordinates in real time; calculating the target line-of-sight yaw angle and target line-of-sight pitch angle required for the camera's optical axis to align with the target based on the target's spatial position point and the current global positioning coordinates; and sending the target line-of-sight yaw angle and target line-of-sight pitch angle to the photoelectric gimbal for tracking and focusing.

[0091] Considering that traditional low-level flight controllers fix the initial position dead zone radius when performing hovering tasks, in strong winds, if this dead zone limitation is not expanded, the control inner loop will force the rotor motors to fight against the airflow, thus hindering the fuselage from gliding towards the target coordinates in the downwind direction. Furthermore, when the fuselage performs translational retreat and active yaw, the relative position between the fuselage and the target to be inspected continuously changes; simply relying on the fixed angle cancellation of the open loop will cause the camera to deviate from the target. Therefore, in the flight control embodiment of this invention, the actual translational distance to avoid the target is added to the initial position dead zone radius to widen the control tolerance, allowing the fuselage to retreat smoothly. Simultaneously, this embodiment of the invention uses the target's spatial position point and the fuselage's real-time global positioning coordinates to recalculate the target line-of-sight yaw angle and the target line-of-sight pitch angle in real time. This design logic, which combines tolerance relaxation with global coordinate parallax calculation, utilizes the overall backward movement and rotation of the camera body to share the limited rotation workload inside the gimbal. It effectively overcomes the risk of crash caused by strict hovering and the problem of missing the target due to line-of-sight deviation, and achieves linkage control between gimbal anti-obstruction and camera stable focusing in gust weather.

[0092] In summary, an adaptive flight control system for power line inspection drones in adverse weather conditions determines the gimbal's angle limit polyhedron by projecting three-dimensional boundaries in the local coordinate system of the drone, intuitively quantifying the physical limit boundaries of gimbal operation. Based on the drone's windward tilt vector, a retreat path is divided, and the expected attitude of each test node is deduced, constructing a camera line-of-sight sweep polyhedron containing local relative lines of sight. This visualizes the dynamic observation lines of sight during the retreat process as spatial entities in advance. Furthermore, spatial interference calculations are performed on the aforementioned gimbal angle limit polyhedron and camera line-of-sight sweep polyhedron to extract direction vector components and determine the interference direction reference angle. This transforms the abstract angle exceeding the limit into a proactive three-dimensional spatial collision warning and accurately locks the future conflict location. Finally, based on the drone's windward tilt vector and interference direction reference angle, proactive flight control is performed in advance. The predicted interference location guides the drone's motion compensation, effectively overcoming the shortcomings of existing technologies that rely on triggering limit exceeding warnings for passive flight control due to adaptive response lag. This results in higher reliability of flight control in adverse weather conditions.

[0093] This invention also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the computer device can execute the aforementioned adaptive flight control system for a power line inspection drone in adverse weather conditions.

Claims

1. An adaptive flight control system for power line inspection drones in adverse weather conditions, characterized in that, The system includes: The data acquisition module is used to acquire the maximum allowable yaw angle and maximum allowable pitch angle of the UAV-borne optoelectronic gimbal, as well as the current pitch angle and current roll angle of the UAV at the current moment, during the UAV's inspection of the target to be inspected, and to establish a local coordinate system of the UAV with the UAV as the origin. The first determining module is used to perform spatial orientation analysis based on the current pitch angle and the current roll angle to determine the UAV's windward tilt vector at the current moment; and to perform three-dimensional boundary projection based on the maximum allowable yaw angle and the maximum allowable pitch angle in the local coordinate system of the aircraft body to determine the gimbal rotation limit polyhedron. The second determining module is used to divide the retreat path based on the UAV's windward tilt vector and determine at least two test nodes; to perform expected attitude deduction based on the current pitch angle, current roll angle and position information of each test node and determine the local relative line-of-sight vector corresponding to each test node; and to perform spatial closure processing based on the start and end points of the local relative line-of-sight vectors of all test nodes to construct a camera line-of-sight sweep polyhedron located in the local coordinate system of the aircraft. The UAV flight control module is used to perform spatial interference calculations and extract directional vector components based on the gimbal rotation limit polyhedron and the camera line-of-sight sweep polyhedron to determine the interference direction reference angle; and to perform flight control of the power inspection UAV based on the UAV's windward tilt vector and the interference direction reference angle.

2. The adaptive flight control system for power line inspection drones in adverse weather conditions according to claim 1, characterized in that, The process of obtaining the UAV's windward tilt vector includes: Based on the sine value of the current pitch angle, determine the X-axis component of the UAV's windward tilt vector; The Y-axis component of the UAV's windward tilt vector is determined based on the product of the sine of the current roll angle and the cosine of the current pitch angle. By zeroing the data, the Z-axis elevation component of the UAV's windward tilt vector is determined. A three-dimensional vector is synthesized based on the X-axis direction component, Y-axis direction component, and Z-axis elevation component to generate the UAV's windward tilt vector.

3. The adaptive flight control system for power line inspection drones in adverse weather conditions according to claim 1, characterized in that, The process of obtaining the gimbal rotation limit polyhedron includes: Obtain the reference unit vector directly in front of the UAV in the local coordinate system of the UAV; construct a two-dimensional angle boundary based on the maximum allowable yaw angle and the maximum allowable pitch angle; perform discrete sampling on the two-dimensional angle boundary according to a preset angle sampling interval to determine at least two extreme yaw angles; perform matrix rotation operation on the reference unit vector directly in front based on each extreme yaw angle to determine at least two boundary rays; project the spatial endpoints of each boundary ray based on a preset unit spherical radius constant to determine the discrete spatial endpoints corresponding to each boundary ray; perform spatial convex hull closure processing based on all discrete spatial endpoints to determine the gimbal rotation angle limit polyhedron.

4. The adaptive flight control system for power line inspection drones in adverse weather conditions according to claim 1, characterized in that, The process of acquiring the test node includes: Obtain the downwind retreat unit vector corresponding to the opposite direction of the UAV's windward tilt vector; obtain the UAV's initial spatial position at the current moment; determine the single-step spatial displacement vector based on the product between the preset single-step sliding length scalar and the downwind retreat unit vector; and determine all test nodes by successively superimposing the position coordinates according to the preset maximum total number of test nodes based on the initial spatial position point and the single-step spatial displacement vector.

5. The adaptive flight control system for power line inspection drones in adverse weather conditions according to claim 4, characterized in that, The process of obtaining the local relative line-of-sight vector includes: Based on the Euclidean distance between the UAV's initial spatial position and each test node at the current moment, the cumulative slip distance of each test node is determined; the cumulative slip distance is negatively correlated and mapped to determine the attitude decay weight. The current pitch angle is weighted based on the attitude attenuation weight to determine the expected pitch angle of each test node; the current roll angle is weighted based on the attitude attenuation weight to determine the expected roll angle of each test node. Obtain the target spatial location point of the target to be inspected; use the vector formed by the position coordinates of each test node pointing to the target spatial location point as the global absolute line-of-sight vector corresponding to each test node; construct the inverse pose transformation matrix mapping each test node to the local coordinate system of the aircraft based on the expected pitch angle and expected roll angle of each test node; multiply the global absolute line-of-sight vector and the inverse pose transformation matrix to determine the initial relative line-of-sight vector; multiply the unit vector of the initial relative line-of-sight vector with a preset unit spherical radius constant to determine the local relative line-of-sight vector corresponding to each test node.

6. The adaptive flight control system for power line inspection drones in adverse weather conditions according to claim 5, characterized in that, The process of acquiring the polyhedron by the camera's line of sight includes: In the local coordinate system of the machine body, starting from the origin of the local coordinate system, the local line-of-sight endpoint of each test node is determined based on the local relative line-of-sight vector corresponding to each test node; and a spatial convex hull closure process is performed based on all local line-of-sight endpoints to construct a camera line-of-sight sweep polyhedron.

7. The adaptive flight control system for power line inspection drones in adverse weather conditions according to claim 4, characterized in that, The process of obtaining the reference angle of the interference direction includes: The gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron are input into the GJK collision interference detection algorithm, which outputs the spatial boundary distance between the gimbal corner limit polyhedron and the camera line-of-sight sweep polyhedron, as well as the nearest boundary interference point. In the local coordinate system of the aircraft, when the spatial boundary spacing is greater than zero, the vector formed by pointing from the origin of the local coordinate system to the nearest boundary interference point is taken as the spatial interference direction vector; when the spatial boundary spacing is equal to zero, the downwind retreat unit vector is taken as the spatial interference direction vector. The X-axis projection component and the corresponding Y-axis projection component of the spatial interference direction vector are extracted and subjected to bivariate arctangent operation to determine the reference angle of the interference direction.

8. The adaptive flight control system for a power line inspection drone in adverse weather conditions according to claim 4, characterized in that, The process of controlling the power line inspection drone's flight based on the drone's windward tilt vector and the interference direction reference angle includes: The distance to the target for translational avoidance is determined by multiplying the preset single-step sliding length scalar with the preset maximum total number of measured nodes; the three-dimensional displacement vector is determined by multiplying the downwind retreat unit vector with the distance to the target for translational avoidance; and the coordinates of the target for translational avoidance in the downwind direction are determined by superimposing the coordinates of the initial spatial position point and the three-dimensional displacement vector. Based on the product of the interference direction reference angle and the preset yaw feedforward compensation ratio gain, the fuselage yaw compensation rotation angle is determined; the fuselage yaw compensation rotation angle is superimposed with the current yaw orientation of the UAV to determine the nose control orientation command; based on the downwind translation target coordinates and the nose control orientation command, a cooperative compensation flight path command is generated to perform flight control of the power line inspection UAV based on the cooperative compensation flight path command.

9. The adaptive flight control system for a power line inspection drone in adverse weather conditions according to claim 3, characterized in that, The preset angle sampling interval ranges from 1 degree to 5 degrees.

10. The adaptive flight control system for a power line inspection drone in adverse weather conditions according to claim 4, characterized in that, The maximum number of preset measurement nodes is set to 10.

Images