Unmanned aerial vehicle attitude control method for complex scene cruise

Abstract

The invention discloses an unmanned aerial vehicle attitude control method for complex scene cruise, and belongs to the technical field of unmanned aerial vehicle attitude control. The invention aims to solve the problems that an existing unmanned aerial vehicle attitude control method is insufficient in robustness under complex environment disturbance, unreasonable in safety redundancy distribution and lack of foresight and adaptive ability. Therefore, the invention provides an unmanned aerial vehicle attitude control method, and the method comprises the steps: dynamically up-regulating or down-regulating the long-term disturbance baseline and risk level of each grid and the neighborhood thereof through the consistency deviation between the attitude error standard deviation in a certain time window and the expected error of a corresponding gear, enabling the disturbance atlas and risk distribution to be continuously self-corrected along with the flight data, and achieving the unmanned aerial vehicle attitude control. Therefore, a real environment can still be accurately reflected after long-term operation, risk identification misjudgment and control strategy mismatching caused by aging or distortion of an environment model are reduced, and attitude control robustness and overall task completion efficiency of the unmanned aerial vehicle during cruising in a complex scene are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) attitude control technology, and more specifically, to a method for UAV attitude control for cruising in complex scenarios. Background Technology

[0002] Currently, attitude control for UAVs used in complex scenarios such as urban canyons, high-rise building clusters, valley turbulence, and rain and snow still mostly relies on pre-tuned fixed-gain controllers or parameter adaptation under limited operating conditions. Existing solutions typically treat the flight environment as a time-varying one-dimensional disturbance sequence, lacking fine-grained raster modeling of the three-dimensional spatial domain. They also fail to maintain disturbance statistics and risk levels in each spatial grid over the long term, and it is even more difficult to uniformly map factors such as attitude control margin, mission turn sensitivity, and actuator health status into risk indicators that can be directly used by the control layer. During long-range cruise, aircraft frequently switch between different disturbance environments such as building-around zones, open areas, and rain / snow areas. If the control parameters are still tuned according to a single operating condition, it is easy to encounter problems such as insufficient attitude margin or even divergence in high-disturbance zones, while being overly conservative and inefficient in low-disturbance zones. At the same time, the distribution of environmental disturbances will change during long-term operation. Existing methods lack a mechanism for closed-loop self-correction of environmental disturbance models and risk levels using operational data such as attitude errors, causing the environmental model to gradually deviate from the actual operating conditions, and the safety redundancy configuration to become increasingly mismatched with the actual environment.

[0003] In response to the above situation, the specific technical problem to be solved by this invention is that when performing long-range cruise missions in complex environmental disturbance scenarios, existing UAV attitude control lacks a closed-loop mechanism that uses an airspace grid as a carrier to uniformly model local disturbance statistics, attitude control margin, mission sensitivity, and actuator health status, and dynamically reconfigures control parameter levels accordingly, while using attitude error statistics to make online corrections to disturbance maps and risk levels. This results in control parameters responding lagging to environmental changes, and disturbance models are prone to distortion after long-term operation. Consequently, in high-risk areas such as building bypass and rain / snow adhesion, there is insufficient attitude control margin or increased oscillation, while in low-risk areas, there is a waste of safety redundancy, making it difficult to balance flight safety and mission efficiency throughout the entire flight range. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide an attitude control method for unmanned aerial vehicles (UAVs) used for cruising in complex scenarios, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for UAV attitude control during cruise in complex scenarios includes the following steps: The mission airspace is divided into three-dimensional grids to form a grid set. For each grid, the short-term disturbance mean, long-term disturbance baseline, disturbance category identifier, risk level, control parameter level for attitude control, and expected attitude error model are initialized. During flight, attitude, control variables, and acceleration are acquired periodically. The disturbance intensity is calculated based on the attitude error, control increment, and acceleration modulus. The current position is mapped to the grid to obtain the target grid. The disturbance quantity of the target grid is updated, and the disturbance category of the target grid is determined based on the neighborhood difference. Three types of risk indices are calculated based on disturbance amount and flight status information. The three types of risk indices are classified and the highest level is taken. The risk level and risk coefficient are obtained by combining the disturbance category. The target control parameter level is selected and the attitude control parameters are set according to the forward risk and execution health level. The attitude control output is calculated based on the attitude control parameters. The standard deviation of the attitude error is statistically analyzed within the time window and compared with the expected standard deviation of the error to obtain the consistency deviation. The long-term disturbance baseline and risk level of the grid and its neighborhood are adjusted based on the consistency deviation. The above steps are repeated until the cruise mission is completed.

[0006] In a preferred embodiment, the three-dimensional grid division includes determining the grid size and number of layers according to the horizontal boundary and height of the task airspace, initializing the short-term disturbance mean of each grid to zero, the long-term disturbance baseline to the default baseline value, and the risk level to a unified initial level, and pre-storing the correspondence between each grid and the control parameter level and the expected attitude error model.

[0007] In a preferred embodiment, acquiring attitude, control quantity, and acceleration includes acquiring attitude angle, attitude angle command, control quantity command, and body acceleration vector at a fixed sampling period, defining attitude error as the difference between attitude angle and attitude angle command, defining control increment as the difference between control quantity commands in adjacent sampling periods, and defining disturbance intensity as a linear combination of attitude error weighted norm, control increment weighted norm, and acceleration magnitude according to preset weighting coefficients.

[0008] In a preferred embodiment, updating the target grid perturbation amount includes obtaining the short-term perturbation mean by using a first time window to apply a moving average of the perturbation intensity, and accumulating the long-term perturbation baseline by using an exponentially weighted moving average. When determining the perturbation category based on neighborhood differences, the neighborhood difference degree is obtained by comparing the average of the target grid's long-term perturbation baseline with that of the neighborhood grid's long-term perturbation baseline, and the target grid is classified and labeled accordingly.

[0009] In a preferred embodiment, the three risk indices include an environmental intensity index, a control margin index, and a mission sensitivity index. The environmental intensity index is calculated from the maximum component of the disturbance intensity and wind speed modulus after normalization. The control margin index is calculated from the minimum value of the attitude angle and the remaining angle of the attitude limit. The mission sensitivity index is calculated from a combination function of the trajectory curvature radius, flight altitude, and mission type.

[0010] In a preferred embodiment, classifying the three types of risk indices includes dividing each risk index into at least three levels according to a multi-level threshold. Taking the highest level includes selecting the highest value among the three risk index levels as the aggregate level, and mapping the aggregate level to the disturbance category identifier as a grid risk level. The risk coefficient is calculated by the grid risk level through a monotonic mapping function and stored in the corresponding grid.

[0011] In a preferred embodiment, the forward-looking risk includes selecting several forward-looking track points in the flight controller according to the planned trajectory, mapping the forward-looking track points one by one to the corresponding grid, reading the risk level of each grid and taking the maximum value as the forward-looking risk, and the execution health level includes calculating the root mean square of the difference between the expected acceleration and the actual acceleration within a preset time window and comparing it with a multi-level threshold to obtain a discrete health level.

[0012] In a preferred embodiment, selecting the target control parameter level based on the forward-looking risk and the execution health level includes accessing a preset decision table using the grid risk level, forward-looking risk, and execution health level as indexes to obtain the target control parameter level identifier. When setting the attitude control parameters, the corresponding attitude control parameter set is read according to the target control parameter level, and the attitude control parameters are linearly or piecewise linearly scaled according to the risk coefficient.

[0013] In a preferred embodiment, the statistical attitude error standard deviation includes performing sliding statistics on the attitude error sequence within a second time window to obtain the attitude error standard deviation. The expected error standard deviation is obtained by looking up a table based on the expected attitude error model of the target grid, the current control parameter level, and the grid risk level. The consistency deviation is the difference between the attitude error standard deviation and the expected error standard deviation, and coefficients for updating are generated accordingly.

[0014] In a preferred embodiment, adjusting the long-term disturbance baseline and risk level of the grid and its neighborhood includes selecting the target grid and adjacent grids in the horizontal and vertical directions when the consistency deviation exceeds a preset threshold, converting the consistency deviation into an update amount through a piecewise mapping function, and adjusting the long-term disturbance baseline and risk level of each grid according to the update amount. When executing steps two to four in a loop, the task completion flag is used as the termination condition.

[0015] The technical effects and advantages of the UAV attitude control method for cruising in complex scenarios according to the present invention are as follows: This invention divides the three-dimensional airspace into grids and maintains short-term disturbance mean, long-term disturbance baseline, disturbance category, and risk level within each grid. This enables the generation of a spatially resolved environmental disturbance map after multiple UAV passages through typical complex scenarios such as building bypass zones and rain / snow areas. Based on this, it comprehensively evaluates the environmental intensity index (formed by single-step disturbance intensity and wind speed modulus), the control margin index (derived from the difference between attitude angle and attitude limit), and the task sensitivity index (derived from trajectory curvature, etc.). A comprehensive risk level and continuous risk coefficient are generated by taking the most dangerous factor and superimposing the disturbance category. Furthermore, different risk levels are selected based on the forward path risk and execution health level. The control parameters are adjusted to automatically switch to enhanced stability or highly conservative settings and tighten attitude margin in high-risk grids, while maintaining normal gain in low-risk grids to improve cruise efficiency. At the same time, by utilizing the consistency deviation between the standard deviation of attitude error and the expected error of the corresponding setting within a certain time window, the long-term disturbance baseline and risk level of each grid and its neighborhood are dynamically adjusted up or down. This allows the disturbance map and risk distribution to continuously self-correct with flight data, thus accurately reflecting the real environment even after long-term operation. This reduces the risk identification omissions and control strategy mismatches caused by environmental model aging or distortion, and improves the attitude control robustness and overall mission completion efficiency of the UAV when cruised in complex scenarios. Attached Figure Description

[0016] Figure 1 This is a flowchart of the spatial grid division and parameter initialization process of the present invention; Figure 2 This is a flowchart of the disturbance statistics and grid disturbance category labeling process of the present invention; Figure 3 This is a flowchart of the risk assessment and control parameter selection process for this invention. Figure 4 This is a flowchart of the attitude error consistency analysis and disturbance map self-correction process of the present invention; Figure 5 This is a timing diagram for attitude control of a drone used for cruising in complex scenarios according to the present invention. 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.

[0018] This invention provides an attitude control method for UAVs used in complex scenarios, aiming to address the problems of insufficient robustness, unreasonable allocation of safety redundancy, and lack of foresight and adaptive capabilities in existing UAV attitude control systems under complex environmental disturbances. This method achieves refined perception of environmental disturbances, multi-dimensional risk assessment, forward-looking control parameter reconfiguration, and closed-loop self-correction by constructing a tightly coupled four-layer system.

[0019] Example 1, refer to Figures 1 to 5 This invention provides an attitude control method for a drone used in complex scenarios, comprising the following steps: Step S1: Airspace Grid Division and Parameter Initialization. Before the UAV begins its cruise mission, the three-dimensional airspace is first divided into grids. Each grid represents a specific spatial region. Simultaneously, the disturbance statistics (such as short-term disturbance mean and long-term disturbance baseline), risk level, control parameter settings, and expected error model for each grid are initialized. These initialization parameters can be set based on historical flight data, weather forecasts, or empirical values. For example, the initial short-term disturbance mean and long-term disturbance baseline can be set to zero or a small baseline value, the risk level can be set to the lowest, and the control parameter settings can be set to the normal gain mode.

[0020] Step S2: During flight, data such as attitude, control parameters, acceleration, and wind speed are collected. Key flight data, including but not limited to: Attitude error The current posture Compared with reference attitude The deviation between them.

[0021] Control increment That is, the current control quantity Control quantity compared to the previous moment The changes between them.

[0022] acceleration modulus That is, the current acceleration vector The length of the module.

[0023] Wind speed modulus That is, the current estimated wind speed vector The length of the module.

[0024] Location This refers to the current three-dimensional spatial position of the drone.

[0025] These data are in fixed time periods Data is collected to provide real-time input for subsequent disturbance analysis and risk assessment. It should be noted that the fixed time period can be set according to actual circumstances.

[0026] Step S3: Calculate the single-step disturbance intensity And update the short-term disturbance mean. and long-term disturbance baseline Specifically: In each sampling period, the single-step perturbation intensity at the current moment is first calculated. This strength is determined by a combination of attitude error, control increment, and acceleration modulus, and its calculation formula can be: ;in, These are preset weighting coefficients used to balance the effects of different disturbance components; This is the acceleration due to gravity.

[0027] Next, based on the drone's current location Through mapping function Determine the grid where the drone is located. Then, update the short-term perturbation mean of the raster. and long-term disturbance baseline Short-term disturbance mean Reflecting recent The calculation formula for the observation disturbance within this grid is as follows: ;in, The time window length, The number of observations within this grid. It is an indicator function, when The value is 1 if the condition is met, and 0 otherwise. Long-term disturbance baseline. The exponential moving average method is used to update the grid, reflecting the long-term perturbation trend. Its calculation formula is as follows: ;in, For smoothing coefficients, The current time falls on the grid Internal disturbances If the drone is not currently in the grid Inside, then It remains unchanged.

[0028] Step S4, to distinguish between occasional minor fluctuations and persistently high-disturbance air fields, comprehensively considers both short-term / long-term comparisons in the time dimension and spatial differences between the field itself and its neighborhood. Specifically: First, calculate the grid. Neighborhood Dissimilarity : ;in, To be with grid A set of grid cells that are adjacent in three dimensions.

[0029] Then, the raster is processed using the following combined conditions. Classified into three categories: Category 0: Low-disturbance zone. When hour.

[0030] Category 1: Stable high-disturbance region. When and and This indicates that short-term disturbances are high, relatively high in the long term, and the differences within the neighborhood are small, suggesting that it is a large area of ​​high disturbance.

[0031] Category 2: Disturbance boundary region. When and This indicates a significant difference between this grid and its neighbors, suggesting it may be located on the edge of a strong wind zone, the edge of a building's flow path, etc. Use integer labels. This indicates the categories of these grid perturbations.

[0032] Step S5: Calculate the environmental intensity index Control margin index and task sensitivity index Specifically, this step assesses the current flight risk from three dimensions: 1. Environmental Intensity Index First, normalize the single-step disturbance intensity and wind speed modulus: (when (Updated in real time); then, the environmental strength is defined as the most unfavorable component: ; 2. Control margin index According to the aircraft attitude constraints Define the current margin: The minimum value among the three is taken as the control margin index. ; 3. Task Sensitivity Index Based on the task type and local trajectory geometry, such as the trajectory radius of curvature. ,definition: The smaller the path radius, the sharper the turn. The closer to 1, the more sensitive it is to attitude errors; during cruise straight sections... , .

[0033] Step S6: Based on the most dangerous criterion, the grid risk level is obtained. and risk coefficient Specifically: The above three indices Converted to three levels respectively .For example, The definition of is: ; Similar but with reversed thresholds, meaning smaller margin and higher risk. Similarly, the maximum value is taken as the overall risk level: At the same time, the disturbance category of the previous layer is... This is also included in the comprehensive assessment. Finally, the levels are standardized to a continuous risk coefficient: .

[0034] Step S7: Based on the current risk, the forward path risk, and the execution status, determine the control parameter level. Specifically: This step begins by assessing the health of the implementing agency. In recent times... Within each cycle, compare the expected acceleration. Compared with the actual measured acceleration : ;according to Health levels are implemented in three tiers. : Next, a forward path risk analysis is performed. Based on the future path provided by the planning module... trackpoints Mapped to raster sequence The corresponding risk level can be obtained by looking up the table. Take the maximum value among them as the forward risk: .

[0035] Then, three sets of control parameters are pre-designed: Level A (normal gain, efficiency priority), Level B (enhanced attitude stability, moderately conservative), and Level C (highly conservative mode, safety priority). This is based on the current overall risk level. Forward-looking risks and implement health levels Use a logical decision table to determine the gear. Specifically: First, calculate the total risk. Example of a rule: If ,but ;like ,but ;like ,but Finally, at the corresponding gear level, based on the risk factor... Calculate the specific parameters of PID / LQR / MPC, taking PID as an example: It should be noted that each gear position still allows for adjustments. Fine-tuning is done, but major control parameters are selected by the logic described above.

[0036] Step S8: Calculate the attitude control output at the selected gear and send it to the actuator. Using the updated control parameters from step S7, calculate the attitude controller output. It then sends the data to the drone's actuators to complete the attitude control for this cycle.

[0037] Step S9: Statistically analyze the actual attitude error sequence and its consistency with the expected error model to correct the perturbation map and risk level. Specifically: For the recent The attitude error over each cycle, and the statistical standard deviation. ,in According to the current gear and risk level In ground simulation / experiment, a desired standard deviation of error is pre-calibrated. Define consistency deviation. : ; Comprehensive correction based on consistency deviation: if It is assumed that the spectrum is consistent with the closed-loop behavior and no correction is needed. and This indicates that the actual situation is more dangerous than predicted. The following actions should be taken: Establish the long-term perturbation baseline of this grid. Increase the scale factor by one; this will change the risk level of the grid. Increase by at least one level (saturate to 2); adjust the neighboring raster accordingly. and .like and This indicates that the model is too conservative and can be gradually reduced. Alternatively, the risk level may be reduced after multiple verifications, thereby improving efficiency. These corrections will affect all calculations starting from S3 in the next cycle, forming a closed-loop self-calibration.

[0038] Step S10: Repeat steps S2 to S9 to complete the entire complex scene cruise mission. Throughout the entire cruise mission, the UAV continuously executes steps S2 to S9 in a loop to achieve adaptive attitude control in complex scenes.

[0039] Example 2: This example will further refine the parameter settings in Example 1 and illustrate their application in specific complex scenarios.

[0040] Specifically, the raster division and initialization parameters can be: Airspace grid: Assuming the cruising area is 10km x 10km x 1km, it can be divided into 100x100x10 grids, each grid being 100m x 100m x 100m in size.

[0041] Weighting coefficients: .

[0042] Time window length Corresponding to 10 sampling periods, smoothing coefficient .

[0043] Disturbance category threshold: .

[0044] Normalization parameter: (m / s).

[0045] Attitude restrictions: .

[0046] Task sensitivity reference radius m.

[0047] Risk level threshold: .

[0048] Perform health assessment parameters: This corresponds to 50 sampling periods. Number of forward-looking waypoints .

[0049] Closed-loop calibration parameters: This corresponds to 20 sampling periods. .

[0050] In an alternative example, in a complex wind field scenario, i.e., a building disturbance scenario: Assuming a drone is cruising between tall buildings in a city, where complex building-around-flow conditions cause drastic changes in the local wind field, the method specifically includes the following steps: The drone flies along a preset path and collects data in real time.

[0051] When a drone enters a grid near a tall building, its attitude error is due to the building's flow around it. and control increment It will increase significantly, leading to a decrease in the intensity of single-step disturbances. Increase. Short-term disturbance mean. A rapid rise. If there is a long-term flow around the region, it will cause a long-term disturbance to the baseline. It will also gradually increase. Meanwhile, due to the drastic changes in the wind field at the building edge, the long-term disturbance baseline differences between neighboring grids are significant, leading to increased neighborhood variability. The value is relatively high. At this point, the grid is likely to be labeled as a disturbed boundary region, i.e., category 2.

[0052] In this case, within the disturbed boundary region, the environmental intensity index Increase. When the drone's current attitude is close to its attitude limit, such as when the drone tilts at a large angle to counteract crosswinds, adjust the control margin index. Decrease. When the drone is making a sharp turn, the mission sensitivity index decreases. This increases the risk level. It will be assessed as a higher level, such as level 2, with a risk factor of [missing information]. Close to 1.

[0053] Furthermore, in forward path analysis, if the trackpoint ahead is also found to enter a similar building bypass area, the forward path risk increases. It will also be very high. (Assuming the health level of the implementing agency...) Normal. According to the logical decision table, due to the total risk... At a very high level, such as Level 2, the system will select a highly conservative mode (Level C). In this mode, the gain parameters of the attitude controller (such as a PID controller) will be adjusted, for example, reducing the P gain to reduce over-response to disturbances and increasing the D gain to improve damping, thereby making the UAV attitude more stable, but the response speed may decrease slightly.

[0054] Finally, after flying in highly conservative mode for a period of time, the system will calculate the standard deviation of the actual attitude error. If the actual error is greater than the expected error The large value indicates that the model's risk assessment of the area is still insufficient, and the system will raise the long-term perturbation baseline of that grid. and risk level This influences neighboring grids, allowing the system to adjust control parameters earlier and more conservatively the next time it passes through the region. Conversely, if the actual error is much smaller than the expected error, the model may be too conservative, and the system will slowly reduce relevant parameters to improve efficiency.

[0055] Another alternative example is a scenario where rain or snow causes changes in aerodynamic characteristics and sensor degradation: Assuming the drone is flying in rain or snow, the rain and snow adhering to the wings will alter its aerodynamic characteristics, and the optical sensors and GNSS signals may be interfered with. The specific steps of the method are as follows: Drones fly in rain and snow, collecting data.

[0056] Changes in aerodynamic characteristics caused by rain and snow increase the difficulty of attitude control for drones and lead to attitude errors. and control increment Increase the intensity of single-step disturbance Increase. Since rain and snow are regional phenomena, both short-term and long-term disturbance mean values ​​will increase, and the neighborhood difference may be low. The grid may be labeled as a stable high-disturbance area, i.e., category 1.

[0057] In this environment, the environmental intensity index It will be very high. If sensor degradation leads to inaccurate attitude estimation, the control margin index will be high. The sensitivity index may decrease. If the mission requires flight over complex low-altitude terrain, the mission sensitivity index will decrease. It will also be very high. After comprehensive assessment, the grid risk level... It will also be relatively high.

[0058] If the forward-looking path indicates that the drone will continue to fly in rainy or snowy areas, forward-looking path risk... It will also be very high. Furthermore, rain and snow may cause a decline in the performance of actuators, such as motors, resulting in a lower execution status index. Raise the health level. The speed can be reduced, for example, to level 1 or level 2. In this case, the system will select either an enhanced attitude stabilization mode (level B) or a highly conservative mode (level C), adjust control parameters to improve attitude stability, and may reduce flight speed to reduce aerodynamic loads.

[0059] The closed-loop correction mechanism continuously optimizes the disturbance spectrum and risk level based on actual flight performance to ensure the system's safety and robustness in adverse weather conditions.

[0060] As can be seen from the detailed description of the above embodiments, the present invention achieves intelligent and adaptive attitude control of cruise drones in complex scenarios through multi-level comprehensive analysis and closed-loop self-correction mechanism, effectively solving many problems existing in the prior art and significantly improving the flight safety and mission efficiency of drones in complex environments.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An attitude control method for unmanned aerial vehicles (UAVs) used for cruising in complex scenarios, characterized in that, Includes the following steps: The mission airspace is divided into three-dimensional grids to form a grid set. For each grid, the short-term disturbance mean, long-term disturbance baseline, disturbance category identifier, risk level, control parameter level for attitude control, and expected attitude error model are initialized. During flight, attitude, control variables, and acceleration are acquired periodically. The disturbance intensity is calculated based on the attitude error, control increment, and acceleration modulus. The current position is mapped to the grid to obtain the target grid. The disturbance quantity of the target grid is updated, and the disturbance category of the target grid is determined based on the neighborhood difference. Three types of risk indices are calculated based on disturbance amount and flight status information. The three types of risk indices are classified and the highest level is taken. The risk level and risk coefficient are obtained by combining the disturbance category. The target control parameter level is selected and the attitude control parameters are set according to the forward risk and execution health level. The attitude control output is calculated based on the attitude control parameters. The standard deviation of the attitude error is statistically analyzed within the time window and compared with the expected standard deviation of the error to obtain the consistency deviation. The long-term disturbance baseline and risk level of the grid and its neighborhood are adjusted based on the consistency deviation. The above steps are repeated until the cruise mission is completed.

2. The UAV attitude control method for cruising in complex scenarios according to claim 1, characterized in that: The three-dimensional grid division includes determining the grid size and number of layers according to the horizontal boundary and height of the mission airspace, initializing the short-term disturbance mean of each grid to zero, the long-term disturbance baseline to the default baseline value, and the risk level to a unified initial level, and pre-storing the correspondence between each grid and the control parameter level and the expected attitude error model.

3. The UAV attitude control method for cruising in complex scenarios according to claim 1, characterized in that: Acquiring attitude, control quantity, and acceleration includes collecting attitude angle, attitude angle command, control quantity command, and body acceleration vector at a fixed sampling period. The attitude error is defined as the difference between the attitude angle and the attitude angle command, the control increment is defined as the difference between the control quantity commands in adjacent sampling periods, and the disturbance intensity is defined as a linear combination of the attitude error weighted norm, the control increment weighted norm, and the acceleration magnitude according to preset weighting coefficients.

4. The UAV attitude control method for cruising in complex scenarios according to claim 1, characterized in that... ； Updating the target grid perturbation includes obtaining the short-term perturbation mean by using a first time window to move the perturbation intensity as a moving average, and accumulating the long-term perturbation baseline by using an exponentially weighted moving average. When determining the perturbation category based on neighborhood differences, the neighborhood difference degree is obtained by comparing the average value of the target grid's long-term perturbation baseline with that of the neighborhood grid's long-term perturbation baseline, and the target grid is classified and labeled accordingly.

5. The UAV attitude control method for cruising in complex scenarios according to claim 1, characterized in that: The three types of risk indices include the environmental intensity index, the control margin index, and the mission sensitivity index. The environmental intensity index is calculated from the maximum component of the disturbance intensity and wind speed modulus after normalization. The control margin index is calculated from the minimum value of the attitude angle and the remaining angle of the attitude limit. The mission sensitivity index is calculated from a combination function of the trajectory curvature radius, flight altitude, and mission type.

6. The UAV attitude control method for cruising in complex scenarios according to claim 5, characterized in that, The classification of the three risk indices includes dividing each risk index into at least three levels according to a multi-level threshold. The maximum level is selected by choosing the highest value among the three risk index levels as the aggregate level, and the aggregate level is mapped to the disturbance category identifier as a grid risk level. The risk coefficient is calculated by the grid risk level through a monotonic mapping function and stored in the corresponding grid.

7. The UAV attitude control method for cruising in complex scenarios according to claim 6, characterized in that, The forward-looking risk includes selecting several forward-looking track points in the flight controller according to the planned trajectory, mapping the forward-looking track points one by one to the corresponding grid, reading the risk level of each grid and taking the maximum value as the forward-looking risk. The execution health level includes calculating the root mean square of the difference between the expected acceleration and the actual acceleration within a preset time window and comparing it with multi-level thresholds to obtain a discrete health level.

8. The UAV attitude control method for cruising in complex scenarios according to claim 7, characterized in that, Selecting the target control parameter level based on forward-looking risk and execution health level includes accessing a preset decision table using grid risk level, forward-looking risk, and execution health level as indexes to obtain the target control parameter level identifier. When setting attitude control parameters, the corresponding attitude control parameter set is read according to the target control parameter level, and the attitude control parameters are linearly or piecewise linearly scaled according to the risk coefficient.

9. The UAV attitude control method for cruising in complex scenarios according to claim 1, characterized in that, The statistical attitude error standard deviation is obtained by performing sliding statistics on the attitude error sequence within the second time window. The expected error standard deviation is obtained by looking up a table based on the expected attitude error model of the target grid, the current control parameter level, and the grid risk level. The consistency deviation is the difference between the attitude error standard deviation and the expected error standard deviation, and coefficients for updating are generated accordingly.

10. The UAV attitude control method for cruising in complex scenarios according to claim 9, characterized in that, Adjusting the long-term disturbance baseline and risk level of the grid and its neighborhood includes selecting the target grid and its adjacent grids in the horizontal and vertical directions when the consistency deviation exceeds a preset threshold, converting the consistency deviation into an update amount through a piecewise mapping function, and adjusting the long-term disturbance baseline and risk level of each grid according to the update amount. When executing steps two to four in a loop, the task completion flag is used as the termination condition.

Images