Unmanned aerial vehicle attitude feedforward compensation control method for building wind field disturbance

Abstract

The invention discloses an unmanned aerial vehicle attitude feed-forward compensation control method for building wind field disturbance, and belongs to the technical field of unmanned aerial vehicle flight control, and the method comprises the steps: building a mapping relation between a building space position and a wind field disturbance characteristic; processing the real-time wind field data in combination with the mapping relation to generate a wind field prediction result; calculating a feed-forward compensation amount, and synchronously determining a dynamic weight factor reflecting the disturbance intensity of the wind field; obtaining an attitude error between a current attitude and an expected attitude of the unmanned aerial vehicle, and calculating a feedback control quantity; and performing weighted fusion on the feedforward compensation quantity and the feedback control quantity by using the dynamic weight factor, and generating a final control instruction to control the attitude of the unmanned aerial vehicle. According to the method, the wind field mapping relation is established and the wind field is predicted to calculate the feed-forward compensation quantity, and then the dynamic weight factor is used to carry out weighted fusion on the feed-forward compensation quantity and the feedback control quantity, so that the disturbance of the wind field can be actively inhibited, and the attitude control stability and precision of the unmanned aerial vehicle in a complex environment are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) flight control technology, and in particular to a UAV attitude feedforward compensation control method for building wind field disturbances. Background Technology

[0002] Unmanned aerial vehicles (UAVs), commonly known as drones, are unmanned aircraft controlled by radio remote control equipment or their own program control devices. The attitude control system is the core of ensuring stable flight, maintaining or changing the aircraft's roll, pitch, and yaw attitude by adjusting the outputs of various actuators. With the increasingly widespread application of UAVs in urban inspection, logistics delivery, and emergency rescue, their ability to fly stably in complex urban canyon environments with towering buildings is crucial.

[0003] Currently, attitude control of unmanned aerial vehicles (UAVs) commonly employs feedback control methods based on attitude error, with the proportional-integral-derivative (PID) controller being the most widely used technology. This type of method acquires the UAV's current attitude using sensors such as an onboard inertial measurement unit (IMU), compares it with the desired attitude to obtain the attitude error, and then calculates the corresponding control input based on this error to drive the motors to adjust their speeds and correct the attitude. This control method is simple in structure, easy to implement, and achieves good control performance in open environments with relatively stable wind conditions.

[0004] However, in densely built-up urban environments, buildings can severely obstruct, accelerate, and separate airflow, creating extremely complex eddies, gusts, and turbulence. Traditional feedback control methods are essentially "post-hoc" adjustments, meaning the controller can only respond based on detected errors after wind disturbances have already caused attitude deviations. Due to the inherent delay in the control system's response, when a UAV encounters sudden and severe gusts, its attitude often has already fluctuated significantly, leading to flight trajectory deviations and decreased stability, posing safety hazards. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method for UAV attitude feedforward compensation control in response to building wind field disturbances. This method involves establishing a building wind field mapping relationship and predicting the wind field to calculate the feedforward compensation amount. Then, a dynamic weighting factor is used to weight and fuse the feedforward compensation amount and the feedback control amount. This approach can proactively suppress wind field disturbances and improve the stability and accuracy of UAV attitude control in complex environments.

[0006] The above objectives can be achieved through the following approach: A method for UAV attitude feedforward compensation control for building wind field disturbances includes: acquiring building structure data and establishing a mapping relationship between building spatial location and wind field disturbance characteristics; acquiring real-time wind field data and processing the real-time wind field data in conjunction with the mapping relationship to generate wind field prediction results; calculating feedforward compensation amount based on the wind field prediction results and simultaneously determining a dynamic weighting factor reflecting the intensity of wind field disturbances; acquiring the attitude error between the current attitude and the desired attitude of the UAV and calculating the feedback control amount; and using the dynamic weighting factor to weight and fuse the feedforward compensation amount and the feedback control amount to generate a final control command to control the UAV attitude.

[0007] Optionally, establishing the mapping relationship between building spatial location and wind field disturbance characteristics includes: analyzing building structure data, extracting building height, shape and layout, and generating building structure parameters; performing fluid dynamics analysis based on the building structure parameters to generate an initial wind field disturbance mode containing eddy current and accelerated wind characteristics; collecting historical wind field data to train the initial wind field disturbance mode and establishing the mapping relationship between building spatial location and wind field disturbance characteristics.

[0008] Optionally, acquiring real-time wind field data includes: measuring wind speed and direction using airborne sensors to acquire raw wind field data; acquiring UAV motion state data provided by the navigation system; and using the UAV motion state data to perform motion compensation on the raw wind field data to eliminate the influence of the UAV's own motion and generate real-time wind field data.

[0009] Optionally, generating the wind field prediction result includes: performing time-series analysis on the real-time wind field data to obtain the current wind field trend characterizing the rate of change of wind speed and the rate of change of wind direction; obtaining the predicted location of the UAV in the next short period of time, and extracting spatial disturbance information corresponding to the predicted location from the mapping relationship; and correcting the current wind field trend by combining the spatial disturbance information to generate the wind field prediction result.

[0010] Optionally, the calculation of the feedforward compensation includes: establishing a moment balance equation characterizing the aerodynamic characteristics of the UAV, inputting the wind field prediction result into the moment balance equation, calculating the disturbance moment caused by the predicted wind field on the UAV body; and solving the control input used to counteract the disturbance based on the disturbance moment, as the feedforward compensation.

[0011] Optionally, determining the dynamic weighting factor reflecting the intensity of wind field disturbance includes: analyzing the amplitude and rate of change of the wind field prediction results, quantifying the degree of wind field turbulence, and generating a disturbance intensity assessment value; inputting the disturbance intensity assessment value into a weight mapping function to generate a dynamic weighting factor.

[0012] Optionally, the calculation of the feedback control quantity includes: measuring the current attitude angle of the UAV through an attitude sensor and performing a difference calculation with the desired attitude angle to generate an attitude error; performing a proportional calculation on the attitude error to obtain a proportional control component; performing an integral calculation on the attitude error to obtain an integral control component; performing a differential calculation on the attitude error to obtain a differential control component; and combining the proportional control component, the integral control component, and the differential control component to generate a feedback control quantity.

[0013] Optionally, the method further includes: acquiring raw wind field measurements in real time and performing motion compensation in conjunction with the current motion state of the UAV to obtain measured wind field data; comparing the measured wind field data with the wind field prediction results at the corresponding time and location to calculate the prediction deviation; and adjusting the mapping relationship between the building's spatial location and wind field disturbance characteristics using an online optimization algorithm based on the prediction deviation to reduce the deviation of subsequent predictions.

[0014] Optionally, generating the final control command includes: multiplying the feedforward compensation amount and the feedback control amount by the feedforward weight coefficient and the feedback weight coefficient in the dynamic weight factor, respectively, and performing a weighted summation to generate a fused control torque; and performing control allocation calculation on the fused control torque according to the UAV rotor layout to generate the final control command.

[0015] Based on the same inventive concept, this invention also provides a UAV attitude feedforward compensation control system for building wind field disturbances. The system includes: a building wind field modeling module for acquiring building structure data and establishing a mapping relationship between building spatial location and wind field disturbance characteristics; a wind field prediction module for acquiring real-time wind field data and processing the real-time wind field data in conjunction with the mapping relationship to generate a wind field prediction result; a feedforward compensation calculation module for calculating the feedforward compensation amount based on the wind field prediction result and simultaneously determining a dynamic weighting factor reflecting the intensity of wind field disturbance; an attitude error processing module for acquiring the attitude error between the current attitude and the desired attitude of the UAV and calculating the feedback control amount; and a fusion control module for using the dynamic weighting factor to weight and fuse the feedforward compensation amount and the feedback control amount to generate a final control command to control the attitude of the UAV.

[0016] Compared with the prior art, the present invention has the following advantages: This invention establishes a mapping relationship between building spatial location and wind field disturbance characteristics, and combines real-time wind field data to predict future wind fields, achieving proactive compensation for wind field disturbances. This method can calculate the corresponding compensation control quantity before wind field disturbances actually affect the attitude of the UAV, thereby actively offsetting them, suppressing attitude deviations caused by disturbances, and improving the flight stability of the UAV in complex building environments.

[0017] This invention proposes a dynamic weighted fusion control strategy based on the intensity of wind field disturbances. By evaluating and predicting the turbulence level of the wind field in real time, the system can dynamically adjust the weights of feedforward compensation and feedback control in the final control command. This adaptive control authority allocation mechanism enables the control system to adopt the optimal control strategy when facing wind field disturbances of different intensities, enhancing the system's robustness and adaptability in variable environments and ensuring a smooth transition and efficient response in control performance.

[0018] This invention incorporates an online optimization mechanism for mapping relationships, enabling wind field models to possess self-learning and self-correction capabilities. By continuously comparing in-flight measured wind field data with the model's predictions and adjusting the mapping relationships in real time based on prediction deviations, the accuracy of the wind field model can be continuously improved. This mechanism not only compensates for the deficiencies in initial modeling but also allows the model to adapt to dynamic environmental changes, thereby providing more reliable prior information for feedforward control and forming a virtuous cycle of continuous performance optimization.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for UAV attitude feedforward compensation control for building wind field disturbances according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the building wind field mapping relationship generation process according to an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of wind field prediction according to an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the structure of a UAV attitude feedforward compensation control system for building wind field disturbance according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes a method for UAV attitude feedforward compensation control for building wind field disturbances. It adopts the method of establishing a building wind field mapping relationship and predicting the wind field to calculate the feedforward compensation amount, and then using a dynamic weight factor to weight and fuse the feedforward compensation amount and the feedback control amount. This method can actively suppress wind field disturbances and improve the stability and accuracy of UAV attitude control in complex environments.

[0027] The method described in this embodiment specifically includes: Obtain building structure data and establish a mapping relationship between building spatial location and wind field disturbance characteristics; Real-time wind field data is acquired and processed in conjunction with the mapping relationship to generate wind field prediction results. Based on the wind field prediction results, the feedforward compensation amount is calculated, and the dynamic weighting factor reflecting the intensity of wind field disturbance is determined simultaneously. Obtain the attitude error between the current attitude and the desired attitude of the UAV, and calculate the feedback control quantity; The dynamic weighting factor is used to weight and fuse the feedforward compensation amount and the feedback control amount to generate the final control command to control the attitude of the UAV.

[0028] Specifically, the method first analyzes building structure data to establish a priori building wind field model, which reveals the intrinsic relationship between specific spatial locations and wind field disturbance characteristics. During actual UAV flight, this method combines real-time measured wind field data with the prior model, enabling not only perception of current wind conditions but also prediction of wind fields in the near future. Based on this prediction, a feedforward compensation quantity is calculated in advance. This control quantity aims to proactively counteract the upcoming wind field disturbance torque. Simultaneously, a traditional feedback control loop operates in parallel, calculating the feedback control quantity by measuring the deviation between the UAV's actual and desired attitudes. This is used to correct all unpredicted disturbances and model errors. This invention does not simply add the feedforward and feedback control quantities; instead, it dynamically generates a weighting factor based on the intensity of the wind field prediction results, and uses this factor to weight and fuse the feedforward compensation and feedback control quantities. The stronger the wind field disturbance, the greater the weight of the feedforward compensation; conversely, the greater the weight of the feedback control, ultimately generating an optimal fused control command that adapts to environmental changes. This invention introduces a feedforward compensation mechanism based on wind field prediction to actively suppress wind field disturbances. It applies a counter-torque before the disturbance actually has an impact, weakening the wind's effect on the UAV's attitude. The introduction of a dynamic weighting factor gives the control system intelligent adaptive capabilities, allowing it to flexibly adjust the focus of the control strategy according to the severity of the wind field disturbance. In environments with strong disturbances, it decisively employs feedforward-driven strong suppression, while in environments with weak disturbances, it relies on feedback control for fine-tuning. This cooperative control strategy enables the UAV to perform tasks more smoothly and reliably, avoiding the risk of attitude instability caused by sudden gusts or eddies, and enhancing the safety and reliability of UAVs operating in urban environments.

[0029] Optionally, establishing the mapping relationship between building spatial location and wind field disturbance characteristics includes: Analyze building structure data, extract building height, shape and layout, and generate building structure parameters; Based on the building structure parameters, a fluid dynamics analysis is performed to generate an initial wind field disturbance mode that includes eddy current and accelerated wind characteristics. Historical wind field data is collected to train the initial wind field disturbance pattern, establishing a mapping relationship between building spatial location and wind field disturbance characteristics.

[0030] Specifically, to establish the mapping relationship between building spatial location and wind field disturbance characteristics, the target building environment first needs to be digitally represented. High-precision building structure data is obtained by analyzing Building Information Modeling (BIM), 3D Geographic Information System (GIS) data, or LiDAR scanning data. Subsequently, geometric analysis is performed on this data to extract building structure parameters that significantly affect the wind field. These parameters include, but are not limited to, the precise 3D coordinates, height, facade shape of each building, and the relative spacing and arrangement of the building group. These parameters together constitute a digital building scene, laying the foundation for subsequent analysis. Next, based on the generated building structure parameters, Computational Fluid Dynamics (CFD) simulation analysis is conducted. A virtual computational domain containing the building model is constructed, and reasonable boundary conditions are set to define the incoming wind speed profiles at different heights. By solving the Navier-Stokes equations, the complex flow phenomena generated when air flows through the building group are simulated, thereby generating a 3D meshed initial wind field disturbance model. This model can characterize typical wind field disturbances such as leeward eddies caused by building shading effects and accelerated winds generated by building slot effects, clarifying the theoretical distribution of wind speed and direction at different spatial locations. However, pure theoretical simulations deviate from the actual environment, thus requiring correction. Long-term historical wind field data is collected by deploying ground-based anemometers in the target area or using wind speed sensors on UAVs to conduct patrols under different weather conditions. This data contains wind field information for specific spatial locations in real-world environments. Finally, the collected historical wind field data is used as training samples to correct and optimize the initial wind field disturbance model. Machine learning algorithms, such as neural networks or Gaussian process regression, are employed, using the simulated wind field data as prior knowledge. By learning the differences between historical data and simulation results, the model parameters are continuously adjusted, ultimately establishing a mapping relationship from building spatial location to wind field disturbance characteristics. Figure 2 As shown, after data training, the initial wind field disturbance pattern is corrected, forming a revised mapping relationship that more accurately reflects the complexity of the real environment. This method integrates the theoretical prediction capability of fluid dynamics simulation with the real-world correction effect of historical data. It not only overcomes the shortcomings of single simulation methods that do not match the actual environment, but also solves the problem that measured data alone cannot cover the entire three-dimensional space. The final generated mapping relationship provides the UAV with a high-precision "wind field map," enabling it to predict the type and intensity of wind field disturbances it will encounter based on its own position. This provides crucial prior information for achieving precise feedforward compensation control, thereby improving the stability and safety of the UAV flying in complex building environments.

[0031] Optionally, acquiring real-time wind field data includes: Wind speed and direction are measured using airborne sensors to obtain raw wind field data; Acquire the drone's own motion status data provided by the navigation system; The original wind field data is compensated using the drone's own motion data to eliminate the influence of the drone's own motion and generate real-time wind field data.

[0032] Specifically, airborne sensors integrated into the UAV's fuselage, such as porous probes or ultrasonic anemometers, are used to measure the airflow velocity and direction relative to the UAV in real time. This measurement yields raw wind field data, which is essentially a vector superposition of the actual atmospheric wind and the airflow generated by the UAV's own motion. Therefore, this data does not directly reflect wind field disturbances in the external environment. To separate the actual atmospheric wind, it is necessary to simultaneously acquire the UAV's own motion state data provided by the navigation system. This navigation system is typically a high-precision integrated navigation system that combines inertial measurement unit (IMU) and Global Navigation Satellite System (GNSS) information, providing the UAV's real-time velocity vector, position, and attitude angles in a ground reference coordinate system. After acquiring these two sets of data, motion compensation calculations are performed to remove the component contributed by the UAV's own motion from the raw wind field data. This process is based on the vector relationship of relative motion and can be characterized by the following formula: , in, This represents the real-time wind field data generated after motion compensation, i.e., the actual atmospheric wind speed vector in the ground reference coordinate system; This represents the raw wind field data measured by airborne sensors in the airframe coordinate system, i.e., the apparent wind speed vector; This represents the drone's own motion state data provided by the navigation system, i.e., the drone's velocity vector relative to the ground; It is an attitude transformation matrix, calculated based on the real-time attitude angles provided by the navigation system, used to transform the apparent wind speed vector in the body coordinate system. The system is rotated and transformed to a ground reference coordinate system. By performing this vector operation, the influence of the UAV's own motion on wind speed measurement is eliminated, thus obtaining real-time wind field data that reflects the true disturbances of the external environment. This method solves the core problem of airborne sensor measurements being contaminated by the aircraft's own motion, decomposing the mixed relative wind measurements into pure environmental wind disturbances. This wind field data is a prerequisite for subsequent wind field prediction and feedforward compensation control, ensuring that the environmental information relied upon by the control system is objective and true, rather than artifacts caused by its own maneuvering. This improves the UAV's ability to perceive external wind field disturbances and lays a data foundation for attitude control.

[0033] Optionally, the generated wind field prediction results include: Time-series analysis of the real-time wind field data yields the current wind field trend, representing the rate of change of wind speed and the rate of change of wind direction. Obtain the predicted location of the drone in the near future, and extract the spatial perturbation information corresponding to the predicted location from the mapping relationship; The current wind field trend is corrected by combining the spatial disturbance information to generate a wind field prediction result.

[0034] Specifically, firstly, time-series analysis is performed on continuously acquired real-time wind field data, caching a short time window sequence of real-time wind field data. Algorithms such as numerical difference or Kalman filtering are then used to calculate the rate of change of this sequence at the current moment, thus obtaining the current wind field trend characterizing the rate of change of wind speed and wind direction. Simultaneously, based on the UAV's flight mission planning and current motion state, its predicted location within a short future prediction time domain is calculated through kinematic extrapolation. This predicted location is then used as a query index, input into a pre-established mapping relationship between building spatial locations and wind field disturbance characteristics, thereby extracting specific spatial disturbance information corresponding to that future location. This spatial disturbance information is based on deterministic wind field characteristics generated by the building structure, such as eddy current intensity or accelerated wind effects at specific locations. Finally, to generate the final wind field prediction result, the above two types of information are dynamically fused. Through a correction process, the current wind field trend obtained from time-series analysis is combined with the spatial disturbance information extracted from the mapping relationship. This combination process can be expressed by the following equation: , in, It is the wind field prediction result, that is, the predicted future wind speed vector; It is a predicted wind speed vector obtained by extrapolating time based on the current wind field trend. It is calculated by using the current real-time wind field data and its rate of change. It is spatial disturbance information extracted from the mapping relationship that corresponds to the predicted location of the UAV, representing the quasi-static wind field characteristics of that location; It is a fusion weighting coefficient whose value can be dynamically adjusted according to flight status to balance the immediacy of time trends with the certainty of spatial information. For example... Figure 3 As shown, the left half of the curve represents the collected real-time wind field data. Starting from the current moment, the dashed line represents the current wind field trend, which predicts that the wind speed will change steadily. However, based on the predicted location, the drone finds from the wind field map that it will enter a wind field anomaly area at some point in the future; this is the spatial disturbance information correction. Superimposing this spatial information onto the time trend yields the final wind field prediction result. This method transcends the limitations of relying solely on historical data extrapolation or static wind field maps, achieving a composite prediction of future wind fields by combining the temporal dynamics and spatial structure of the wind field.

[0035] Optionally, the calculation of the feedforward compensation includes: A moment balance equation characterizing the aerodynamic characteristics of the UAV is established. The wind field prediction results are input into the moment balance equation to calculate the disturbance torque caused by the predicted wind field on the UAV body. The control input used to counteract the disturbance is calculated in reverse based on the disturbance torque, and used as the feedforward compensation amount.

[0036] Specifically, the first step is to establish a mathematical model describing the aerodynamic characteristics of the UAV. The core of this model is the torque balance equation. This equation quantifies the relationship between the external torque acting on the UAV body and the UAV's angular acceleration. During execution, the predicted wind field results are used as key inputs and substituted into this dynamic model to calculate the disturbance torque caused by the predicted wind field on the UAV body. This calculation process can be expressed by the following formula: , in, It is the calculated disturbance moment vector, whose components correspond to the roll, pitch and yaw axes in the UAV body coordinate system; This represents a nonlinear function, the specific form of which is determined by the aerodynamic model of the UAV; It is the wind field prediction result obtained, that is, the predicted relative wind speed vector; This is the aerodynamic moment coefficient matrix of the UAV, a crucial parameter characterizing its aerodynamic properties. This coefficient matrix is ​​typically determined beforehand through methods such as wind tunnel testing, computational fluid dynamics simulation, or flight data identification. It establishes a quantitative relationship between the relative wind speed vector and the resulting aerodynamic moment. Using this formula, it's possible to predict the rotational impact of an upcoming wind disturbance on the UAV's attitude. After obtaining the disturbance moment... Then, a reverse solution is performed to determine a control input that can completely counteract this disturbance. Ideally, the applied control torque... It should be equal in magnitude and opposite in direction to the disturbance torque, that is... This calculation This is the final output feedforward compensation, which exists in the form of control torque commands and directly represents the active compensation operation required to maintain attitude stability. Unlike traditional feedback control, which corrects only after attitude deviations occur, this method, by establishing a dynamic model and combining it with wind field prediction, can calculate the required compensation control amount before wind disturbances actually act on the UAV and cause attitude errors. This proactive control strategy can shorten the system's response delay and weaken the impact of wind disturbances on the UAV's attitude. As a result, the UAV's flight attitude is more stable, and even when encountering sudden gusts or entering complex vortex regions, it can suppress violent attitude fluctuations, improving flight stability and control accuracy.

[0037] Optionally, the determination of the dynamic weighting factor reflecting the intensity of wind field disturbance includes: The amplitude and rate of change of the wind field prediction results are analyzed to quantify the degree of wind field turbulence and generate a disturbance intensity assessment value. The disturbance intensity assessment value is input into the weight mapping function to generate a dynamic weight factor.

[0038] Specifically, to determine the dynamic weighting factor reflecting the intensity of wind field disturbances, this method first performs in-depth analysis of the wind field prediction results to assess the severity of environmental disturbances in real time. This process includes two aspects: first, analyzing the amplitude of the wind field prediction results, i.e., the magnitude of the predicted wind speed, which directly reflects the baseline level of disturbance energy; and second, analyzing its rate of change, i.e., how quickly the predicted wind speed changes over time, which characterizes the turbulence or gust characteristics of the wind field. By combining these two dimensions, the degree of wind field turbulence can be comprehensively quantified. In practice, the amplitude of the wind field prediction results is calculated, and its time derivative or difference operation is performed to obtain the rate of change. Subsequently, these two indicators are fused through a pre-defined evaluation function to generate a single scalar value, i.e., the disturbance intensity evaluation value. This evaluation function is a weighted combination designed to comprehensively reflect the combined impact of wind speed magnitude and the severity of its change on the attitude stability of the UAV. Finally, this disturbance intensity evaluation value is used as an independent variable and input into a pre-designed weighting mapping function. The weighting mapping function is a nonlinear function, such as an sigmoid function, that smoothly maps a wide range of disturbance intensity assessment values ​​to a standardized interval, typically between 0 and 1. Specifically, a sigmoid function can be used. Its key characteristic is the ability to flexibly set the sensitivity range and response speed of the weights to disturbance intensity by adjusting its slope and center point parameters, ensuring a smooth transition between feedforward and feedback control permissions under different disturbance levels and avoiding abrupt changes in control commands. The output of this function is the dynamic weighting factor. When the disturbance intensity assessment value is low, the weighting factor output by the function is correspondingly small; while when the disturbance intensity assessment value increases, the output weighting factor also increases, eventually tending towards a saturation upper limit. This intelligent allocation of control permissions enhances the robustness and adaptability of the control system in variable wind field environments, ensuring that the UAV achieves optimal attitude control performance regardless of the disturbance intensity.

[0039] Optionally, the calculated feedback control quantity includes: The attitude sensor measures the current attitude angle of the UAV and calculates the difference between it and the desired attitude angle to generate the attitude error. The attitude error is proportionally calculated to obtain the proportional control component; The attitude error is integrated to obtain the integral control component; The attitude error is differentiated to obtain the differential control component; The proportional control component, integral control component, and derivative control component are combined to generate a feedback control quantity.

[0040] Specifically, this method employs a classic proportional-integral-derivative (PID) control strategy for closed-loop attitude regulation of the UAV. First, the current attitude angles of the UAV are measured in real time using onboard attitude sensors, such as an inertial measurement unit (IMU), typically expressed as roll, pitch, and yaw angles. Simultaneously, the desired attitude angles are obtained from the mission planning module of the flight control system or remote control commands. Then, the difference between the current and desired attitude angles on the corresponding axes is calculated to generate a three-dimensional attitude error vector, which describes the degree and direction of deviation between the current and target attitudes. Next, the PID controller processes this attitude error in parallel. The proportional gain stage multiplies the attitude error by a proportional gain coefficient to obtain the proportional control component, which provides an instantaneous corrective torque proportional to the current error. The integral gain stage integrates the cumulative value of the attitude error over time and multiplies it by an integral gain coefficient to obtain the integral control component. The differential operation calculates the rate of change of attitude error and multiplies it by a differential gain coefficient to obtain the differential control component. This component predicts the trend of error change and provides a damping torque to suppress excessively rapid attitude changes, preventing overshoot and oscillation. Finally, the three independently calculated proportional control component, integral control component, and differential control component are vector-added together to generate the final feedback control quantity. This process can be expressed by the following formula: , in, It is a feedback control quantity, which is a control torque vector; The attitude error vector, which varies with time, is obtained by subtracting the current attitude angle from the desired attitude angle. , and These are the proportional, integral, and differential gain coefficient matrices, which are pre-tuned according to the UAV's dynamic characteristics to achieve optimal control performance. This method enables the feedback control system to suppress attitude deviations caused by various unmodeled disturbances, sensor noise, and model parameter biases, ensuring that the UAV can stably track the desired attitude. This is the foundation for ensuring the stable operation of the entire control system.

[0041] Optionally, the method further includes: The raw wind field measurements are collected in real time, and motion compensation is performed based on the current motion state of the UAV to obtain the measured wind field data. The measured wind field data is compared with the wind field prediction results at the corresponding time and location, and the prediction deviation is calculated. Based on the prediction deviation, an online optimization algorithm is used to adjust the mapping relationship between the building's spatial location and the wind field disturbance characteristics, thereby reducing the deviation in subsequent predictions.

[0042] Specifically, to achieve online optimization of the mapping relationship between building spatial location and wind field disturbance characteristics, this method establishes a closed-loop process of continuous learning and self-correction. During the UAV's flight mission, two operations are performed in parallel. First, raw wind field measurements are continuously collected in real time via onboard sensors and immediately combined with the UAV's current motion state data obtained from the navigation system for motion compensation. The processing method is completely consistent with the aforementioned steps for acquiring real-time wind field data. The output of this process is the measured wind field data at the UAV's current position and time, which can be considered as the "label" of the true wind field at that point. Second, the internal wind field prediction module is invoked to obtain the wind field prediction result generated by the current mapping relationship at the exact same position and time. Subsequently, these two are directly compared by vector to calculate the prediction deviation. This deviation can be defined by the following formula: , in, It is the prediction bias vector, which quantifies the difference between the predicted value and the true value of the current mapping model; These are measured wind field data obtained after motion compensation; This is the wind field prediction result output at the corresponding time and location based on the mapping relationship. The prediction bias is obtained. Immediately afterwards, an online optimization algorithm is initiated. This algorithm, such as recursive least squares or gradient descent-based adaptive algorithms, uses the prediction bias as an error signal input. Based on this error signal, the algorithm calculates the adjustment amount for the mapping relationship model parameters, with the adjustment direction aimed at reducing the prediction bias calculated in this instance. In this way, every valid wind field data collected by the UAV during flight is used as a training sample, fine-tuning and updating the mapping relationship between building spatial location and wind field disturbance characteristics in real time and iteratively, thereby continuously reducing the bias of subsequent predictions during mission execution. This method not only compensates for inherent errors caused by insufficient data or inaccurate simulation during initial modeling, but more importantly, it enables the model to adapt to unforeseen dynamic changes in the environment, such as airflow disturbances caused by other moving objects or short-term changes in weather conditions. Ultimately, through continuous self-improvement, the accuracy of the mapping relationship is continuously enhanced, thereby providing more accurate prior information for feedforward compensation control and further enhancing the stability and reliability of UAV flight in complex building environments.

[0043] Optionally, the generation of final control instructions includes: The feedforward compensation amount and the feedback control amount are multiplied by the feedforward weight coefficient and the feedback weight coefficient in the dynamic weight factor, respectively, and then weighted and summed to generate the fused control torque. The fusion control torque is calculated and distributed based on the UAV rotor layout to generate the final control command.

[0044] Specifically, this process first generates a total control torque that integrates the two control strategies. The calculated feedforward compensation and feedback control quantities are then weighted and summed using a defined dynamic weighting factor. Specifically, the dynamic weighting factor includes a feedforward weighting coefficient and a feedback weighting coefficient, which are typically normalized so that their sum is 1 to ensure the stability of the total control authority. The weighted fusion process can be expressed by the following formula: , In this formula, It is the final generated fused control torque vector, which represents the optimal attitude control intention at the current moment; It is a feedforward compensation amount calculated based on wind field prediction results, which is a control torque vector; It is a feedback control quantity calculated based on attitude error, and it is also a control torque vector; These are feedforward weighting coefficients. These are feedback weighting coefficients, which together constitute the dynamic weighting factor, whose value is dynamically determined based on the wind field disturbance intensity. This is used to obtain the fused control torque. Then, the control is allocated to the various actuators of the UAV, namely the rotors. This step is called control allocation calculation, the core of which is to solve inversely based on the dynamic relationship determined by the inherent rotor layout of the UAV to determine the rotational speed or thrust of each rotor required to generate the fused control torque. This calculation process is usually completed through the inverse operation of a control allocation matrix, which combines the fused control torque of the three axes and a total thrust command from the altitude / position controller to calculate the target rotational speed required for each rotor motor. These target rotational speed signals constitute the final control command and are sent to the electronic speed controllers of each motor for execution. This method achieves seamless integration of intelligent fusion of control strategies and physical execution. The weighted fusion makes the UAV control system highly adaptable. It can find the optimal balance between active "wind-following" feedforward control and passive "bias correction" feedback control according to the actual situation of the environmental wind field, making the control response both rapid and stable. The subsequent control allocation calculation decomposes this comprehensive control intention into instructions to the underlying hardware, ensuring that the decisions of the upper-level algorithm can be executed by the UAV physical system.

[0045] Based on the same inventive concept, such as Figure 4As shown, the present invention also provides a UAV attitude feedforward compensation control system for building wind field disturbances, the system comprising: The building wind field modeling module is used to acquire building structure data and establish a mapping relationship between building spatial location and wind field disturbance characteristics; The wind field prediction module is used to acquire real-time wind field data and process the real-time wind field data in conjunction with the mapping relationship to generate wind field prediction results. The feedforward compensation calculation module is used to calculate the feedforward compensation amount based on the wind field prediction results, and simultaneously determine the dynamic weighting factor that reflects the intensity of wind field disturbance. The attitude error processing module is used to obtain the attitude error between the current attitude and the desired attitude of the UAV and to calculate the feedback control quantity. The fusion control module is used to use the dynamic weighting factor to weight and fuse the feedforward compensation amount and the feedback control amount to generate the final control command to control the attitude of the UAV.

[0046] To verify the feasibility of this invention in practice, it was applied to a drone-based curtain wall inspection task of a super high-rise building in a downtown CBD area. The building has a complex structure and is surrounded by dense buildings, creating a dynamic urban canyon wind field. Traditional drone attitude control methods are difficult to cope with sudden gusts and eddies, resulting in shaky inspection images, unstable flight attitude, and even safety risks.

[0047] To verify the effectiveness of the present invention, an inspection drone equipped with the control system of the present invention was deployed as the experimental group, and another drone of the same model equipped with only a traditional PID feedback controller was used as the control group. The same inspection route was executed under similar weather conditions, and the mission cycle was one month.

[0048] In this embodiment, structural parameters such as building height, facade shape, and building spacing were first extracted by analyzing the Building Information Model (BIM) data of the Shanghai World Financial Center and its adjacent buildings. Based on these parameters, Computational Fluid Dynamics (CFD) simulations were performed to generate an initial wind field disturbance model that included features such as eddies in the leeward zone of the buildings and slotted wind acceleration. Subsequently, this model was trained using wind field data collected from historical flight surveys, ultimately establishing a high-precision mapping relationship between building spatial locations and wind field disturbance characteristics, forming a three-dimensional "wind field map."

[0049] During an inspection mission one afternoon, the experimental drone flew near the western corner of a building. Its onboard ultrasonic anemometer measured raw wind field data, while the navigation system provided the drone's own motion status. The system calculated real-time wind field data for the current location through motion compensation. Based on time-series analysis of this real-time wind field data, the system obtained the current wind field trend. Simultaneously, based on the drone's predicted flight path, the system extracted spatial disturbance information at the corner from the "wind field map," indicating a strong vortex area due to building obstruction. The system combined the current wind field trend with this spatial disturbance information to generate a wind field prediction, predicting that the drone would encounter a lateral gust with wind speeds suddenly increasing from 5 m / s to 12 m / s within the next 1.5 seconds.

[0050] Based on the wind field prediction results, the feedforward compensation calculation module uses the UAV aerodynamic model to quickly calculate the disturbance torque required to counteract the gust, and then solves for the feedforward compensation amount in reverse. Simultaneously, the system analysis shows that the amplitude and rate of change of the predicted wind field are extremely large, and the disturbance intensity assessment value is very high. Therefore, the feedforward weight coefficient in the dynamic weighting factor is adjusted accordingly. Set to 0.8, feedback weight coefficient The value was set to 0.2 accordingly. At this moment, since the attitude had not yet deviated, the PID feedback control quantity was close to zero. The fusion control module weighted and fused the feedforward compensation quantity and the feedback control quantity to generate the final control command. This command drove the rotor to adjust its speed just before the gust arrived, causing the drone to tilt slightly in the direction of the wind. When the gust actually arrived, the drone's attitude only fluctuated slightly by less than 0.5 degrees, while the control group drone experienced a violent sway of more than 5 degrees at the same position, causing the inspection and filming to be interrupted.

[0051] Furthermore, this invention also achieves online optimization of the mapping relationship during flight. For example, when first flying over the western corner, the system compares the measured wind field data with the predicted results and calculates a prediction deviation of 2.5 m / s. The online optimization algorithm immediately adjusts the mapping relationship parameters for that area based on this deviation. After several flight missions, the prediction deviation at this location has been reduced to within 0.8 m / s, improving the prediction accuracy of subsequent flights.

[0052] Data comparison shows that the control method of this invention has advantages in flight stability, environmental adaptability, and control accuracy. Throughout the test period, the average attitude angle deviation of the experimental group UAVs was reduced by approximately 75% compared to the control group; the improvement in attitude stability was particularly significant in key areas encountering strong gusts. Through dynamic weight adjustment, the system achieves the fusion of feedforward and feedback control, ensuring efficient response under different disturbance intensities. The online optimization mechanism ensures the continuous evolution of the wind field model, enabling it to adapt to subtle changes in the real environment.

[0053] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0054] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for UAV attitude feedforward compensation control for building wind field disturbances, characterized in that, The method includes: Obtain building structure data and establish a mapping relationship between building spatial location and wind field disturbance characteristics; Real-time wind field data is acquired and processed in conjunction with the mapping relationship to generate wind field prediction results. Based on the wind field prediction results, the feedforward compensation amount is calculated, and the dynamic weighting factor reflecting the intensity of wind field disturbance is determined simultaneously. Obtain the attitude error between the current attitude and the desired attitude of the UAV, and calculate the feedback control quantity; The dynamic weighting factor is used to weight and fuse the feedforward compensation amount and the feedback control amount to generate the final control command to control the attitude of the UAV.

2. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 1, characterized in that, The establishment of the mapping relationship between building spatial location and wind field disturbance characteristics includes: Analyze building structure data, extract building height, shape and layout, and generate building structure parameters; Based on the building structure parameters, a fluid dynamics analysis is performed to generate an initial wind field disturbance mode that includes eddy current and accelerated wind characteristics. Historical wind field data is collected to train the initial wind field disturbance pattern, establishing a mapping relationship between building spatial location and wind field disturbance characteristics.

3. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 1, characterized in that, The acquisition of real-time wind field data includes: Wind speed and direction are measured using airborne sensors to obtain raw wind field data; Acquire the drone's own motion status data provided by the navigation system; The original wind field data is compensated by using the drone's own motion state data to eliminate the influence of the drone's own motion and generate real-time wind field data.

4. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 3, characterized in that, The generated wind field prediction results include: Time-series analysis is performed on the real-time wind field data to obtain the current wind field trend, which characterizes the rate of change of wind speed and the rate of change of wind direction; Obtain the predicted location of the drone in the near future, and extract the spatial perturbation information corresponding to the predicted location from the mapping relationship; The current wind field trend is corrected by combining the spatial disturbance information to generate a wind field prediction result.

5. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 4, characterized in that, The calculated feedforward compensation includes: A moment balance equation characterizing the aerodynamic characteristics of the UAV is established. The wind field prediction results are input into the moment balance equation to calculate the disturbance torque caused by the predicted wind field on the UAV body. The control input used to counteract the disturbance is calculated in reverse based on the disturbance torque, and used as the feedforward compensation amount.

6. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 4, characterized in that, The dynamic weighting factors that determine the intensity of wind field disturbances include: The amplitude and rate of change of the wind field prediction results are analyzed to quantify the degree of wind field turbulence and generate a disturbance intensity assessment value. The disturbance intensity assessment value is input into the weight mapping function to generate a dynamic weight factor.

7. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 1, characterized in that, The calculated feedback control quantity includes: The attitude sensor measures the current attitude angle of the UAV and calculates the difference between it and the desired attitude angle to generate the attitude error. The attitude error is proportionally calculated to obtain the proportional control component; The attitude error is integrated to obtain the integral control component; The attitude error is differentiated to obtain the differential control component; The proportional control component, integral control component, and derivative control component are combined to generate a feedback control quantity.

8. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 1, characterized in that, The method further includes: The raw wind field measurements are collected in real time, and motion compensation is performed based on the current motion state of the UAV to obtain the measured wind field data. The measured wind field data is compared with the wind field prediction results at the corresponding time and location, and the prediction deviation is calculated. Based on the prediction deviation, an online optimization algorithm is used to adjust the mapping relationship between the building's spatial location and the wind field disturbance characteristics, thereby reducing the deviation in subsequent predictions.

9. The method for UAV attitude feedforward compensation control for building wind field disturbance as described in claim 1, characterized in that, The generation of the final control command includes: The feedforward compensation amount and the feedback control amount are multiplied by the feedforward weight coefficient and the feedback weight coefficient in the dynamic weight factor, respectively, and then weighted and summed to generate the fused control torque. The fusion control torque is calculated and distributed based on the UAV rotor layout to generate the final control command.

10. A UAV attitude feedforward compensation control system for building wind field disturbances, applied to the UAV attitude feedforward compensation control method for building wind field disturbances as described in any one of claims 1-9, characterized in that, The system includes: The building wind field modeling module is used to acquire building structure data and establish a mapping relationship between building spatial location and wind field disturbance characteristics; The wind field prediction module is used to acquire real-time wind field data and process the real-time wind field data in conjunction with the mapping relationship to generate wind field prediction results. The feedforward compensation calculation module is used to calculate the feedforward compensation amount based on the wind field prediction results, and simultaneously determine the dynamic weighting factor that reflects the intensity of wind field disturbance. The attitude error processing module is used to obtain the attitude error between the current attitude and the desired attitude of the UAV and to calculate the feedback control quantity. The fusion control module is used to use the dynamic weighting factor to weight and fuse the feedforward compensation amount and the feedback control amount to generate the final control command to control the attitude of the UAV.

Images