Urban management multi-unmanned aerial vehicle inspection intelligent posture regulation and data linkage management system

Abstract

This invention relates to the field of UAV inspection technology, specifically to an intelligent attitude control and data linkage management system for multi-UAV inspection in urban management. The invention stores the preset mapping relationship between inspection task types and target attitude parameters through a task attitude mapping unit and associates attitude vectors with waypoints. It also utilizes an online data quality assessment unit to perform multi-dimensional quality assessment on inspection image frames and generate attitude control trigger signals. Furthermore, an attitude deviation direct calculation unit calculates the deviation between the current attitude and the target attitude into attitude correction commands. An adaptive control mode switching unit uses standard gain during normal flight segments and switches to high-precision hold mode within critical acquisition windows. Finally, a multi-UAV collaborative attitude compensation unit triggers nearby UAVs for assessment and re-acquisition when single-UAV attitude correction fails. This data quality closed-loop drives attitude control and multi-UAV collaborative compensation, improving the integrity of inspection data in complex urban environments.

Description

Technical Field

[0001] This invention relates to the field of drone inspection technology, and more specifically, to an intelligent attitude control and data linkage management system for multi-drone inspection in urban management. Background Technology

[0002] Drone inspection is an important technology that is widely used in scenarios such as urban facade inspection, bridge under-bridge inspection, and road surface defect identification. Its core lies in obtaining high-quality inspection image data through precise attitude control.

[0003] Existing technologies have explored various aspects of attitude control methods for unmanned aerial vehicles (UAVs). Patent CN202111497209 discloses an adaptive fuzzy PID control method for multi-rotor UAVs, proposing the introduction of an adaptive fuzzy control loop on the basis of classical PID control. When external disturbances occur, online parameter tuning of the PID parameters is performed to improve attitude control performance. CN202511394249 discloses a method, device, medium, and equipment for controlling a variable-load UAV, which processes attitude angle parameters and system gain through inner and outer loop observers to achieve stable attitude control under variable-load conditions. CN20221... 1510054 discloses a time-varying gain active disturbance rejection optimization control method for unmanned helicopters, which designs a time-varying gain extended state observer to estimate and compensate for disturbances in the unmanned helicopter altitude and attitude composite system. CN120580616B discloses a UAV inspection method and system for power transmission lines, which generates flight attitude adjustment strategy through video frame image quality evaluation results to realize the linkage between image quality feedback and attitude control. In addition, an adaptive flight control method based on LSTM deep learning has also been proposed, which trains the model with historical flight data to output control parameters to adapt to complex inspection environments.

[0004] However, the existing technologies still have the following real problems: First, PID gain adjustment focuses on attitude recovery under external disturbances and does not take the quality of inspection data as a direct driving variable for control gain switching. The flight control system lacks scene perception capability for the acquisition window, resulting in the sharing of the same set of control parameters for routine flight and key acquisition, which cannot balance flight efficiency and image acquisition accuracy. Second, the attitude correction capability of a single drone has physical boundaries. When the drone fails to correct its attitude due to hardware limitations or sudden environmental changes, and the acquired data is consistently substandard, the existing technology does not have a closed-loop mechanism for multi-drone collaborative acquisition, resulting in the inability to automatically repair inspection blind spots. Third, current solutions are mostly geared towards single drone models or single mission scenarios, lacking a systematic solution for multi-mission, multi-drone mixed formations in urban management. There is no configurable mapping and linkage relationship between mission type, target attitude, and data quality assessment. To solve this technical problem, we provide an intelligent attitude control and data linkage management system for multi-drone inspection in urban management. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent attitude control and data linkage management system for multi-drone inspection in urban management, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, an intelligent attitude control and data linkage management system for multi-drone inspection in urban management is provided, including:

[0007] The task attitude mapping unit is used to store the preset mapping relationship between urban management inspection task types and target attitude parameters, and to associate the corresponding target attitude parameter vector with each inspection waypoint.

[0008] The online data quality assessment unit is deployed on the UAV's airborne terminal to assess the quality of real-time acquired inspection image frames, generate a data quality score, and generate an attitude control trigger signal when the data quality score is lower than the quality threshold of the corresponding inspection task type.

[0009] The attitude deviation direct calculation unit is used to calculate the deviation between the current actual attitude and the target attitude parameter vector associated with the current waypoint into an attitude correction command based on the attitude control trigger signal.

[0010] The adaptive control mode switching unit is used to adopt standard control gain in the normal flight segment to balance flight efficiency and energy consumption. When it is determined that the current inspection image frame is in the critical acquisition window, the flight control system is switched from the normal mode to the high-precision holding mode to improve the response sensitivity of the control loop and suppress overshoot. After the critical acquisition window ends, it automatically returns to the normal mode.

[0011] The multi-drone collaborative attitude compensation unit is deployed in the cloud. When the data quality score of any drone is continuously lower than the quality threshold and cannot meet the standard even after its own attitude correction, it outputs a quality failure signal and broadcasts the quality failure signal and the corresponding target attitude parameter vector to the neighboring drones. This triggers the neighboring drones to evaluate their own position, remaining endurance and current task load. If intervention is determined, the flight path and attitude of the neighboring drones are adjusted to perform supplementary data collection, and the collected results are transmitted to the data linkage management unit.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0013] The task attitude mapping unit establishes configurable preset mapping relationships between different inspection task types and target attitude parameters. Combined with a digital twin model, a complete attitude parameter vector is pre-set for each waypoint, enabling standardization and structuring of attitude control tasks at the planning level. The online data quality assessment unit calls a dedicated assessment model to perform multi-dimensional quantitative analysis of image frames based on the task type, using the data quality score as a direct criterion for control mode switching. The adaptive control mode switching unit accordingly performs gain scheduling between the normal flight segment and the high-precision hold mode, increasing the proportional and differential coefficients while suppressing integral effects. Within the critical acquisition window, this simultaneously improves attitude response sensitivity and overshoot suppression. The system's control capabilities balance flight efficiency and data acquisition accuracy at the control strategy level. The attitude deviation direct calculation unit converts the attitude angle deviation into actuator correction commands in real time through a proportional control algorithm, forming a fast inner loop that drives attitude correction based on data quality assessment. When single-aircraft attitude correction saturates and fails, the multi-aircraft collaborative attitude compensation unit selects intervenable UAVs in the cloud based on spatial proximity, endurance, and task priority constraints, triggering multi-aircraft collaborative supplementary data acquisition. This transforms the data acquisition blind spots that are inaccessible to single-aircraft control into a distributed compensation control space, eliminating inspection control gaps at the system level and improving the integrity of inspection data and control robustness of multi-aircraft formations in complex urban environments. Attached Figure Description

[0014] Figure 1 This is an overall block diagram of the present invention.

[0015] The meanings of the labels in the diagram are as follows:

[0016] 1. Task attitude mapping unit; 2. Online data quality assessment unit; 3. Direct attitude deviation calculation unit; 4. Control mode adaptive switching unit; 5. Multi-machine collaborative attitude compensation unit; 6. Data linkage management unit. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides an intelligent attitude control and data linkage management system for multi-drone inspections in urban management. Please refer to [link / reference]. Figure 1 As shown, it includes:

[0019] The task attitude mapping unit is used to store the preset mapping relationship between urban management inspection task types and target attitude parameters, and to associate the corresponding target attitude parameter vector with each inspection waypoint.

[0020] The online data quality assessment unit is deployed on the UAV's airborne terminal to assess the quality of real-time acquired inspection image frames, generate a data quality score, and generate an attitude control trigger signal when the data quality score is lower than the quality threshold of the corresponding inspection task type.

[0021] The attitude deviation direct calculation unit is used to calculate the deviation between the current actual attitude and the target attitude parameter vector associated with the current waypoint into an attitude correction command based on the attitude control trigger signal.

[0022] The adaptive control mode switching unit is used to adopt standard control gain in the normal flight segment to balance flight efficiency and energy consumption. When it is determined that the current inspection image frame is in the critical acquisition window, the flight control system is switched from the normal mode to the high-precision holding mode to improve the response sensitivity of the control loop and suppress overshoot. After the critical acquisition window ends, it automatically returns to the normal mode.

[0023] The multi-drone collaborative attitude compensation unit is deployed in the cloud. When the data quality score of any drone is continuously lower than the quality threshold and cannot meet the standard even after its own attitude correction, it outputs a quality failure signal and broadcasts the quality failure signal and the corresponding target attitude parameter vector to the neighboring drones. This triggers the neighboring drones to evaluate their own position, remaining endurance and current task load. If intervention is determined, the flight path and attitude of the neighboring drones are adjusted to perform supplementary data collection, and the collected results are transmitted to the data linkage management unit.

[0024] The preset mapping relationship is for different inspection task types in urban management, including but not limited to facade inspection, bridge bottom inspection, and road surface defect identification. It pre-sets and stores a set of target attitude parameters, including target pitch angle, target yaw angle, target roll angle and target gimbal angle, which match each task type.

[0025] Associating each inspection waypoint with a corresponding target attitude parameter vector, specifically including:

[0026] Based on the pre-planned flight route, and combined with the digital twin city model, each waypoint on the route is assigned a complete target attitude parameter vector that matches the type of mission being executed, retrieved from the preset mapping relationship. The complete target attitude parameter vector is preset during the mission planning stage and bound to the spatial coordinates of the corresponding waypoint, and then sent to the UAV executing the mission.

[0027] The online data quality assessment unit calls the corresponding assessment model based on the specific inspection task type being performed. The assessment model performs multi-dimensional quantitative analysis on the inspection image frame, including sharpness measurement, target integrity check, and illumination uniformity analysis.

[0028] Among them, the sharpness measurement is evaluated by calculating the proportion of high-frequency components in the image; the target integrity check is to calculate the outline sharpness and the proportion in the inspection image for the key target areas preset in the task type; and the illumination uniformity analysis is obtained by evaluating the variance of the overall gray-scale distribution of the inspection image.

[0029] The results of multi-dimensional quantitative analysis are weighted and fused with the weights related to task type to generate a data quality score.

[0030] The attitude deviation direct calculation unit obtains the current actual attitude angle of the UAV in real time, and subtracts the current actual attitude angle from the corresponding target attitude angle in the target attitude parameter vector associated with the current waypoint to obtain the attitude angle deviation including pitch angle deviation, yaw angle deviation and roll angle deviation.

[0031] The calculation process uses a proportional control algorithm to multiply each attitude angle deviation by a preset proportional coefficient to obtain attitude correction commands for driving the UAV flight control actuators corresponding to each attitude channel.

[0032] The conventional flight segment refers to the phase in which the UAV transfers between two preset inspection waypoints in a straight line or a curve.

[0033] The standard control gain refers to a fixed combination of proportional coefficients, integral coefficients, and derivative coefficients used to form a flight control loop in the conventional flight segment.

[0034] In the normal flight phase, the standard control gain means that the flight control system of the UAV uses proportional coefficients, integral coefficients, and derivative coefficients to process attitude deviations and generate flight control commands.

[0035] Determining that the current inspection image frame is within a critical acquisition window includes:

[0036] When the UAV flies to the spatial capture tolerance range of the preset inspection waypoint, a comprehensive judgment is made in combination with the data quality score generated in real time by the online data quality assessment unit. Specifically, when the UAV enters the spatial capture tolerance range and the data quality score does not reach the preset data quality score threshold for the current inspection task type, but has entered the preset quality preparation threshold range that is higher than the data quality score threshold, it is determined that the UAV has entered the critical acquisition window.

[0037] The improvement of the control loop's response sensitivity and suppression of overshoot specifically includes:

[0038] When the UAV's flight control system switches from the conventional mode to the high-precision hold mode, the flight control system increases the proportional coefficient in the standard control gain by a preset amount, introduces a preset differential coefficient increment, and reduces the weight of the integral coefficient. By increasing the proportional coefficient, the response speed to attitude deviation is improved, and by enhancing the differential action, the trend of attitude deviation changes is predicted and suppression force is applied in advance.

[0039] The critical acquisition window is determined to end when the data quality scores of a specified number of inspection image frames continuously output by the online data quality evaluation unit consistently reach or exceed the preset data quality score threshold for the current inspection task type. In this case, the control mode adaptive switching unit automatically restores the flight control system from the high-precision maintenance mode to the normal mode.

[0040] When the cloud receives a substandard quality signal from a drone, it filters out other drones within a preset distance range of the drone that sent the substandard signal as neighboring drones according to the preset proximity judgment rules, and broadcasts an intervention assessment request to the neighboring drones.

[0041] Upon receiving an intervention assessment request, a nearby drone assesses whether the straight-line distance between its current location and the area requiring additional data collection is less than the maximum intervention distance supported by its remaining battery life, and assesses whether the priority of the currently executing task load is lower than a preset intervention task priority threshold. If and only if both distance and priority conditions are met simultaneously, the nearby drone determines itself to be an intervention-enabled drone and sends a confirmation signal to the cloud.

[0042] After receiving a data acquisition instruction from the cloud, which includes the spatial coordinates of the acquisition area and the target attitude parameter vector, the interventionable UAV plans a temporary flight path from its current position to the acquisition area. The temporary flight path follows preset safe flight rules. After arriving at the acquisition area, the interventionable UAV adjusts its attitude to the target state according to the target attitude parameter vector and the attitude correction instruction, and acquires image data of the target area. After the acquisition is completed, the acquired image data, along with its timestamp, spatial location, and acquisition attitude information, is encapsulated and transmitted to the data linkage management unit through a communication link.

[0043] The specific implementation of the intelligent attitude control and data linkage management system for multi-UAV inspections in urban management is as follows: First, the task attitude mapping unit is the foundation for realizing the intelligence of the entire system. It is responsible for storing preset mapping relationships and associating target attitude parameter vectors with waypoints. The preset mapping relationship refers to the pre-setting and storage of a set of target attitude parameters that uniquely matches each inspection task type in urban management, such as facade inspection, bridge under-bridge inspection, and road surface defect identification. This set specifically includes target pitch angle, target yaw angle, target roll angle, and target gimbal angle. The specific process of associating the corresponding target attitude parameter vectors with each inspection waypoint includes two core components: constructing a digital twin city model and generating waypoint attitude vectors based on the model. The core step is the construction of the digital twin city model. This involves integrating geographic information system (GIS) data, oblique photogrammetry data of the urban area, LiDAR scan point cloud data, and building information model (BIM) data of key buildings. This multi-source heterogeneous data is then fused using a coordinate system to generate a 3D mesh model containing the physical appearance and spatial location of entities such as landforms, building structures, roads, and bridges. This model not only represents geometric shapes but also assigns semantic labels to different components through semantic segmentation technology, such as building facades, bridge decks, and asphalt pavements, thus forming a 3D urban infrastructure information model with semantic information. After obtaining this digital twin model, the process of assigning target attitude parameter vectors to waypoints begins. First, based on the specific inspection task, for example… For example, when inspecting the facade of a building, the digital twin model selects the building's exterior surface as the inspection area. Based on preset flight safety distances, sensor field of view angles, and imaging resolution requirements, the mission planning system automatically generates a series of ordered waypoints in the space in front of the inspection area. These waypoints constitute the flight path for this mission. Each waypoint has a three-dimensional spatial position defined by global navigation satellite system coordinates. Next, according to the current mission type, such as facade inspection, the system retrieves a set of target attitude parameters matching the mission type from the preset mapping relationship as a reference template. This template specifies that the gimbal angle should be horizontal (i.e., pitch angle of zero degrees), while the UAV roll angle should remain horizontal. For each specific waypoint, its complete target attitude parameter vector needs to be further refined. This process is accomplished by calculating the relative spatial relationship between the waypoint and the surface of the target to be inspected. Specifically, for each waypoint, the system performs spatial analysis in the digital twin model, calculating the direction vector from the waypoint's location to the center point of the target surface area it is responsible for inspecting. After coordinate transformation, this direction vector is solved into the target yaw angle and target pitch angle in the UAV's body coordinate system, ensuring that the onboard camera's optical axis can be accurately aligned with the center of the target area. At the same time, to ensure that the captured image is level, the target roll angle is set to be parallel to the local horizontal plane, i.e., zero degrees, and the gimbal angle is finely adjusted and compensated in the body coordinate system according to the alignment requirements.Finally, a complete target attitude parameter vector is generated for each waypoint, containing the specific target pitch angle, yaw angle, roll angle, and gimbal angle values ​​calculated using the aforementioned spatial geometric relationships. This vector is calculated during the mission planning phase and bound to the spatial coordinates of the corresponding waypoint, forming a complete waypoint command set containing position and attitude information. Before mission execution, this waypoint command set is transmitted to the UAV performing the mission via a wireless data link. The UAV flight control system and gimbal control system will use this as the reference target for flight and photography.

[0044] Secondly, the online data quality assessment unit is deployed on the UAV's onboard computing equipment. Its function is to perform online quality assessment of real-time acquired images. Based on the specific inspection task being performed by the UAV, such as facade inspection, this unit retrieves a pre-configured evaluation model from the onboard storage unit. The specific construction of this evaluation model is based on the analysis of historical high-quality inspection image samples. The process of achieving multi-dimensional quantitative analysis is as follows: First, sharpness is assessed by calculating the proportion of high-frequency components in the image. Specifically, for each frame of the real-time acquired inspection image, it is first converted from color space to grayscale space to obtain a grayscale image. Then, a spatial gradient operator, such as the Sobel operator, is used to convolve the grayscale image. The process involves calculating the gradient components of each pixel in the image in both the horizontal and vertical directions. The square root of the sum of the squares of the horizontal and vertical gradient components of each pixel is then taken to obtain the gradient magnitude of that pixel. This gradient magnitude characterizes the intensity of local changes in the image at that point, i.e., the intensity of high-frequency information. Subsequently, the average gradient magnitude of all pixels in the entire image is calculated and used as the image frame's sharpness evaluation value. A larger average gradient magnitude indicates more sharp edges and textures in the image, i.e., a higher proportion of high-frequency components, resulting in a sharper image. Secondly, target integrity checking is performed on pre-defined key target areas within the task type. During the task planning phase, the target attitude corresponding to each waypoint is calculated in the digital twin... The model can pre-determine key target areas that should be captured from a given perspective, such as the outline of a window in a facade inspection task, or a suspected crack area in a road surface defect identification task. This process involves projecting a semantically labeled 3D target model from a digital twin model onto a 2D image plane based on the target attitude parameter vector at the current waypoint and camera intrinsics, forming a theoretical target area mask. During real-time evaluation, the evaluation model first processes the acquired inspection image frames using a lightweight semantic segmentation neural network. The neural network outputs pixel areas belonging to the pre-defined key target category, such as windows or cracks, to obtain the actual segmentation mask. The target integrity check includes two sub-items: one is outline sharpness. The illumination uniformity is obtained by calculating the average gradient magnitude of the actual segmentation mask boundaries, similar to the aforementioned sharpness metric, but only for the mask boundary pixels. Secondly, it's determined by the proportion of the target in the image, calculated as the ratio of the total number of pixels in the actual segmentation mask to the total number of pixels in the entire image. Finally, illumination uniformity analysis is obtained by evaluating the variance of the overall grayscale distribution of the inspected image. Specifically, the grayscale image is divided into multiple non-overlapping regular sub-blocks. The average grayscale value of all pixels within each sub-block is calculated, and then the variance of these average grayscale values ​​is calculated. This variance is used as the illumination uniformity evaluation value; the smaller the variance, the closer the average grayscale values ​​of each sub-block are, and the more uniform the overall illumination of the image.After completing the quantitative analysis of four dimensions—sharpness evaluation value, outline sharpness value, target proportion value, and illumination uniformity evaluation value—the evaluation model weights and fuses these values ​​according to the preset weights for the current inspection task type to generate the final data quality score. Specifically, the evaluation value of each dimension is first normalized, mapping it to a numerical range between zero and one. Sharpness evaluation values ​​and outline sharpness values ​​are generally expected to be as high as possible and can be used directly after normalization. The target proportion value has an ideal range; the normalization function ensures it receives the highest score when it falls within this range, decreasing the score if it deviates. The illumination uniformity evaluation value is expected to be as low as possible, and the normalization function performs a reverse processing on it. Finally, weights are assigned to the normalized sharpness values. Weights are assigned to the normalized outline sharpness values. Assign weights to the normalized target proportion values. Assign weights to the normalized illumination uniformity values. These weighting coefficients are preset according to the task type. For example, for a facade inspection task, the weighting is based on the sharpness of the outline. and target proportion weight The score may be set too high because it requires complete and clear identification of building components, which affects the final data quality score. From the formula The calculation shows that, among which This represents the normalized sharpness value. This represents the normalized sharpness value. This represents the normalized target percentage. This represents the normalized illumination uniformity value. The value is a numerical value between zero and one. The higher the score, the better the image quality meets the requirements of subsequent analysis. The generated data quality score will be compared with the quality threshold preset for the current inspection task type. If the score is lower than the threshold, an attitude control trigger signal will be generated.

[0045] The attitude deviation direct calculation unit responds to attitude control trigger signals. Its workflow is as follows: upon receiving a trigger signal from the online data quality assessment unit, it immediately reads the actual attitude angles of the UAV at the current moment from the UAV's inertial measurement unit, including the actual pitch angle, actual yaw angle, and actual roll angle. Simultaneously, it obtains the complete target attitude parameter vector associated with the UAV's current waypoint from the flight management system. This vector contains the target pitch angle, target yaw angle, and target roll angle. The attitude deviation direct calculation unit calculates the difference between the current actual attitude angle and the corresponding target attitude angle one by one. Specifically, it subtracts the actual pitch angle from the target pitch angle to obtain the pitch angle deviation; subtracts the actual yaw angle from the target yaw angle to obtain the yaw angle deviation; and subtracts the actual roll angle from the target roll angle to obtain the roll angle deviation. These deviations reflect the angular differences between the UAV's current attitude and the desired attitude along the three rotational axes. The calculation process employs a proportional control algorithm. This algorithm presets a proportional coefficient for each attitude channel—the pitch channel, yaw channel, and roll channel—denoted as [missing information]. , , Each calculated attitude angle deviation is multiplied by a corresponding preset scaling factor to directly obtain an original control variable, such as the original value of the attitude correction command for the pitch channel. From the formula *The pitch angle deviation is calculated, and similarly, the original control values ​​for the yaw and roll channels can be obtained. and These raw control quantities are numerical values ​​characterizing the control torque or control surface deflection that needs to be applied. To drive the UAV's flight control actuators, such as brushless motors or servo motors, these raw control quantities need to be converted into physical commands that the actuators can recognize. For motors, the command might be the duty cycle of a pulse-width modulation signal adjusting the motor speed; for servos, the command might be the pulse width of the target position. The conversion process is accomplished through a mapping function that linearly maps the raw control quantity U to the safe operating range of the actuator. For example, if the raw control quantity of the pitch channel... The calculation range is from -1 to 1, while the range of the motor control signal is 45% to 55% of the pulse width modulation duty cycle. Therefore, through the linear mapping formula, the final duty cycle = 50% + The final attitude correction command is calculated at 5%, and this final command is sent in real time to the corresponding flight control actuator, thereby driving the UAV to adjust in the direction of reducing attitude deviation.

[0046] The adaptive control mode switching unit manages the control strategy of the UAV in different flight phases. The normal flight phase refers to the phase where the UAV moves between two preset waypoints. This movement typically follows a straight or smooth curved path generated by the flight control system. A straight-line transfer means the UAV flies in a straight line from one waypoint to the next, while a curved transfer means the UAV flies along a curved trajectory generated by a path planning algorithm, taking into account obstacle avoidance and flight smoothness. In the normal flight phase, the system uses a standard control gain, which is a set of fixed coefficients pre-set and embedded in the flight control parameters, including a proportional coefficient. Integral coefficient and differential coefficients These three coefficients together constitute the parameters of the classic proportional-integral-derivative (PID) controller used for attitude control. During normal flight, the flight control system uses this set of standard control gains to construct the flight control loop. Specifically, the flight control system continuously calculates the deviation between the UAV's current position and the target waypoint position, or the deviation between the current attitude and the desired attitude. Then, the PID controller processes these deviations using the standard control gains, with the proportional term... Multiply by the current deviation to provide control action proportional to the deviation; integral term The integral of the deviation is used to eliminate steady-state error; the differential term is... Multiplying by the derivative of the deviation is used to predict the trend of deviation changes and provide damping. The outputs of these three terms are summed to generate the final flight control command used to control the actuator. This set of gain parameters is set to ensure the basic accuracy and stability of flight trajectory tracking without excessively pursuing fast response, thus balancing flight efficiency and energy consumption. When the UAV flies into the spatial acquisition tolerance range of the preset inspection waypoint, the control mode adaptive switching unit begins to intervene. The spatial acquisition tolerance range is a spherical spatial area with a preset radius centered on the target waypoint. When the UAV enters this spherical area, it means that it has approached the target acquisition point. At this time, the unit makes a comprehensive judgment based on the data quality score generated in real time by the online data quality assessment unit. Specifically, the system presets two thresholds, one of which is the quality that must be achieved. Another threshold is a higher quality preparation threshold. When the UAV enters the spatial acquisition tolerance range and the real-time data quality score has not reached the quality threshold, but has entered the quality preparation threshold range, it is determined that it has entered the critical acquisition window. The critical acquisition window is a dynamic time interval during which the system believes that critical data is about to be acquired or is being acquired, requiring higher precision attitude control to ensure data quality. Once the critical acquisition window is determined, the control mode adaptive switching unit immediately switches the flight control system from the conventional mode using standard control gain to the high-precision hold mode. The high-precision hold mode is a control mode specifically designed to achieve more stable and accurate attitude hold within the critical acquisition window. In this mode, the system dynamically adjusts the control gain. Specifically, the proportional coefficient in the standard control gain is adjusted. Increase by a preset amplitude To obtain a new proportionality coefficient At the same time, a preset differential coefficient increment is introduced. To obtain new differential coefficients For integral coefficients If so, then reduce it, for example, set it to Half of the value, or in some cases temporarily set to zero, can be increased by increasing the proportionality coefficient to prevent integral saturation. The flight control system generates larger control commands for the same attitude deviation, thereby improving the control loop's sensitivity to small attitude deviations and enabling the UAV to correct attitude drift more quickly. This is achieved by enhancing the differential coefficients. The system enhances its response to attitude deviation rate of change, which can predict future trends in deviation and output a suppressive control command in advance. This effectively suppresses overshoot and oscillations that may occur during rapid adjustments, allowing the attitude to converge smoothly and accurately to the target value. When the critical acquisition window ends, the system automatically reverts from high-precision hold mode to the standard mode using standard control gain. The condition for determining the end of the critical acquisition window is that the online data quality assessment unit continuously outputs a specified number of inspection image frames, such as five frames, and their data quality scores consistently reach or exceed the preset quality threshold for the current inspection task type. This indicates that the data quality acquired under the current attitude has consistently met the standard, and there is no need to continue maintaining the high-precision control mode. Subsequently, the control mode adaptive switching unit switches the control gain parameter of the flight control system from... , , Switch back to the original , , .

[0047] The multi-drone collaborative attitude compensation unit is deployed on a cloud server to coordinate other drones to supplement data collection when a single drone's self-adjustment is insufficient to meet data quality requirements. This unit is activated when the cloud receives a substandard quality signal uploaded by any drone via the data link. The substandard quality signal includes the drone's identification, current spatial location, persistently low-quality waypoint information, and the target attitude parameter vector associated with that waypoint. The cloud first filters out neighboring drones spatially according to a preset proximity determination rule. This rule typically defines a spherical region with the current location of the drone containing the substandard quality signal as its center and a fixed distance R as its radius. The cloud queries the real-time locations of all online and normal drones, selecting other drones within this spherical region as neighboring drones. Subsequently, the cloud broadcasts an intervention assessment to these neighboring drones. The evaluation request includes the center coordinates of the area requiring re-sampling (i.e., the waypoint coordinates corresponding to the substandard signal) and the target attitude parameter vector of that waypoint. Upon receiving the intervention evaluation request, a nearby UAV initiates its own evaluation process. This evaluation is primarily based on two conditions: First, a distance condition: the UAV calculates the three-dimensional straight-line distance D between its current position and the center coordinates of the area requiring re-sampling. Simultaneously, based on its remaining battery power, average power consumption, and safety margin, the UAV calculates the maximum flight distance S supported by its remaining range. The calculation of the maximum flight distance S must consider the round trip, i.e., the total distance required for the UAV to fly to the re-sampling area and return to its original mission path or safe point. The distance condition is satisfied if and only if the straight-line distance D is less than the maximum flight distance S. Second, a priority condition: the UAV assesses the priority of its currently executing task load. This priority is determined when the task is issued and compared with the preset intervention task priority threshold. Comparison, if and only if Below When the priority condition is met, it indicates that the current task can be temporarily interrupted or delayed, and the supplementary data collection request will be responded to first. Only when both the distance and priority conditions are met simultaneously does the neighboring UAV determine itself as an interferable UAV, and then sends a confirmation signal to the cloud, indicating that it can perform the supplementary data collection task. After receiving the formal supplementary data collection instruction from the cloud, the interferable UAV begins to execute the supplementary data collection operation. The supplementary data collection instruction contains precise spatial coordinates of the supplementary data collection area and target attitude parameter vectors. The spatial coordinates of the supplementary data collection area are directly derived from the waypoint coordinates associated with the substandard signal. The interferable UAV first plans a temporary flight path from its current position to the supplementary data collection area. This path planning must follow preset safe flight rules, including but not limited to maintaining a minimum safe altitude above ground buildings, avoiding known no-fly zones and obstacles, and maintaining a minimum distance from other UAVs. A planning algorithm, such as the A* algorithm or the fast randomized tree search algorithm, is used to search for... A collision-free path from the starting point to the target point, conforming to safe flight rules, allows the UAV to adjust its attitude based on the received target attitude parameter vector after arriving at the supplementary data acquisition area. The adjustment process is consistent with the working principle of the attitude deviation direct calculation unit, which reads its current actual attitude angle, calculates the difference with the target value in the target attitude parameter vector, generates an attitude correction command through a proportional control algorithm, and drives the actuator to move until the deviation between the actual attitude and the target attitude is within the allowable tolerance range. Once the attitude stabilizes at the target state, the UAV controls the onboard imaging equipment to acquire supplementary image data of the target area. After acquisition, the UAV encapsulates the acquired image data, along with the timestamp of the acquisition time, the UAV's own spatial position information at the time of acquisition, and the actual attitude information at the time of acquisition, into a data packet and transmits it to the data linkage management unit via a wireless data link. The unit archives the data, associates it with the original task, and performs subsequent analysis, thereby completing a complete multi-UAV collaborative supplementary data acquisition operation.

[0048] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart attitude control and data linkage management system for multi-drone inspection in urban management, characterized in that: include: The task attitude mapping unit (1) is used to store the preset mapping relationship between the urban management inspection task type and the target attitude parameters, and to associate the corresponding target attitude parameter vector with each inspection waypoint; The online data quality assessment unit (2) is deployed on the UAV airborne terminal to perform quality assessment on the real-time collected inspection image frames to generate a data quality score, and to generate an attitude control trigger signal when the data quality score is lower than the quality threshold of the corresponding inspection task type. The attitude deviation direct calculation unit (3) is used to calculate the deviation between the current actual attitude and the target attitude parameter vector associated with the current waypoint into an attitude correction command based on the attitude control trigger signal; The adaptive control mode switching unit (4) is used to adopt standard control gain in the normal flight segment to balance flight efficiency and energy consumption. When it is determined that the current inspection image frame is in the critical acquisition window, the flight control system is switched from the normal mode to the high precision holding mode to improve the response sensitivity of the control loop and suppress overshoot. After the critical acquisition window ends, it automatically returns to the normal mode. The multi-drone collaborative attitude compensation unit (5) is deployed in the cloud. When the data quality score of any drone is continuously lower than the quality threshold and cannot meet the standard even after its own attitude correction, it outputs a quality failure signal and broadcasts the quality failure signal and the corresponding target attitude parameter vector to the neighboring drones. This triggers the neighboring drones to evaluate their own position, remaining endurance and current task load. When it is determined that intervention is possible, it adjusts the flight path and attitude of the neighboring drones to perform supplementary data collection and transmits the collection results to the data linkage management unit (6).

2. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 1, characterized in that: The preset mapping relationship is for different inspection task types in urban management, including but not limited to facade inspection, bridge bottom inspection, and road surface defect identification. It pre-sets and stores a set of target attitude parameters, including target pitch angle, target yaw angle, target roll angle and target gimbal angle, which match each task type. Associating each inspection waypoint with a corresponding target attitude parameter vector, specifically including: Based on the pre-planned flight route, and combined with the digital twin city model, each waypoint on the route is assigned a complete target attitude parameter vector that matches the type of mission being executed, retrieved from the preset mapping relationship. The complete target attitude parameter vector is preset during the mission planning stage and bound to the spatial coordinates of the corresponding waypoint, and then sent to the UAV executing the mission.

3. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 2, characterized in that: The online data quality assessment unit (2) calls the corresponding assessment model according to the specific inspection task type being performed. The assessment model performs multi-dimensional quantitative analysis on the inspection image frame, including sharpness measurement, target integrity check and illumination uniformity analysis. Among them, the sharpness measurement is evaluated by calculating the proportion of high-frequency components in the image; the target integrity check is to calculate the outline sharpness and the proportion in the inspection image for the key target areas preset in the task type; and the illumination uniformity analysis is obtained by evaluating the variance of the overall gray-scale distribution of the inspection image. The results of multi-dimensional quantitative analysis are weighted and fused with the weights related to task type to generate a data quality score.

4. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 2, characterized in that: The attitude deviation direct calculation unit (3) obtains the current actual attitude angle of the UAV in real time, and subtracts the current actual attitude angle from the corresponding target attitude angle in the target attitude parameter vector associated with the current waypoint to obtain the attitude angle deviation including pitch angle deviation, yaw angle deviation and roll angle deviation. The calculation process uses a proportional control algorithm to multiply each attitude angle deviation by a preset proportional coefficient to obtain attitude correction commands for driving the UAV flight control actuators corresponding to each attitude channel.

5. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 1, characterized in that: The conventional flight segment refers to the phase in which the UAV transfers between two preset inspection waypoints in a straight line or a curve. The standard control gain refers to a fixed combination of proportional coefficients, integral coefficients, and derivative coefficients used to form a flight control loop in the conventional flight segment. In the normal flight phase, the standard control gain means that the flight control system of the UAV uses proportional coefficients, integral coefficients, and derivative coefficients to process attitude deviations and generate flight control commands.

6. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 5, characterized in that: Determining that the current inspection image frame is within a critical acquisition window includes: When the UAV flies to the spatial capture tolerance range of the preset inspection waypoint, a comprehensive judgment is made in combination with the data quality score generated in real time by the data quality online evaluation unit (2). Specifically, when the UAV enters the spatial capture tolerance range and the data quality score does not reach the preset data quality score threshold of the current inspection task type, but has entered the preset quality preparation threshold range that is higher than the data quality score threshold, it is determined that the UAV has entered the critical acquisition window.

7. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 5, characterized in that: The improvement of the control loop's response sensitivity and suppression of overshoot specifically includes: When the UAV's flight control system switches from the conventional mode to the high-precision hold mode, the flight control system increases the proportional coefficient in the standard control gain by a preset amount, introduces a preset differential coefficient increment, and reduces the weight of the integral coefficient. By increasing the proportional coefficient, the response speed to attitude deviation is improved, and by enhancing the differential action, the trend of attitude deviation changes is predicted and suppression force is applied in advance.

8. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 7, characterized in that: The critical acquisition window is determined to end when the data quality scores of the online data quality evaluation unit (2) continuously outputting a specified number of inspection image frames all stably reach or exceed the data quality score threshold preset for the current inspection task type. In this case, the control mode adaptive switching unit (4) will automatically restore the flight control system from the high-precision maintenance mode to the normal mode.

9. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 3, characterized in that: When the cloud receives a substandard quality signal from a drone, it filters out other drones within a preset distance range of the drone that sent the substandard signal as neighboring drones according to the preset proximity judgment rules, and broadcasts an intervention assessment request to the neighboring drones. Upon receiving an intervention assessment request, a nearby drone assesses whether the straight-line distance between its current location and the area requiring additional data collection is less than the maximum intervention distance supported by its remaining battery life, and assesses whether the priority of the currently executing task load is lower than a preset intervention task priority threshold. If and only if both distance and priority conditions are met simultaneously, the nearby drone determines itself to be an intervention-enabled drone and sends a confirmation signal to the cloud.

10. The intelligent attitude control and data linkage management system for multi-UAV inspection in urban management according to claim 9, characterized in that: After receiving a supplementary acquisition instruction from the cloud containing the spatial coordinates of the supplementary acquisition area and the target attitude parameter vector, the interferable UAV plans a temporary flight path from its current position to the supplementary acquisition area. The temporary flight path follows preset safe flight rules. After arriving at the supplementary acquisition area, the interferable UAV adjusts its attitude to the target state according to the target attitude parameter vector and the attitude correction instruction, and performs supplementary image data acquisition of the target area. After the acquisition is completed, the acquired image data, along with its timestamp, spatial location and acquisition attitude information, is encapsulated and transmitted to the data linkage management unit (6) through the communication link.

Images