A method, device and equipment for adaptive imbalance control of unmanned aerial vehicle flight attitude

Abstract

The present invention proposes a method, device and equipment for adaptive imbalance control of the flight attitude of an unmanned aerial vehicle (UAV). The method aims to accurately obtain an ideal control trajectory and identify beneficial imbalance patterns by constructing an inverse trajectory analysis and controllable imbalance strategy generation mechanism, thereby generating a controllable imbalance strategy library. The method also adopts dynamic balance benchmark establishment and trajectory reconstruction technology to construct an imbalance-recovery control trajectory and accurately obtain dynamic response characteristics. The method also introduces imbalance recovery space-time manifold construction and topology analysis technology to extract a core imbalance control feature set through space-time folding transformation and determine key imbalance control parameters. The method also establishes imbalance-recovery mapping rules and an adaptive parameter fusion mechanism to balance control effects and safety requirements, and generate disturbance control parameters and adaptive mapping parameters. The method also combines inverse imbalance joint control and multi-axis collaborative control mechanisms to construct a real-time control output adjustment mechanism to generate flight control signals. Finally, the method forms an adaptive imbalance control system to achieve precise imbalance control of the UAV attitude.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) control technology, and in particular to a method, device and equipment for adaptively controlling the flight attitude of a UAV. Background Art

[0002] As a key application of modern aviation technology, drones are increasingly used across various fields. However, traditional drone attitude control technology primarily relies on error-correction control, treating attitude deviations and imbalances that occur during flight as interference factors that need to be eliminated and using feedback control to restore the attitude to the intended trajectory. While this control method can maintain flight stability, it fails to fully utilize the maneuverability potential inherent in imbalances, limiting the drone's maneuverability and environmental adaptability.

[0003] Existing technologies suffer from the following main problems: a lack of in-depth analysis and effective utilization of imbalance phenomena, making it impossible to identify beneficial imbalance patterns and transform them into control resources; a lack of inverse design methods from ideal trajectory to actual control, making it difficult to establish a precise mapping relationship between imbalance phenomena and control effects; a lack of dynamic balance benchmarks and adaptive control mechanisms, making it impossible to flexibly adjust imbalance control strategies based on changes in flight status; and a lack of a multi-axis coordinated imbalance control system, making it difficult to achieve unified coordination and optimized coordination of imbalance control across all axes. Therefore, there is an urgent need to invent a new UAV attitude control method that can actively identify and utilize imbalance phenomena, establish an adaptive imbalance control strategy, and address the technical issues of existing technologies such as insufficient imbalance utilization, low control efficiency, and poor adaptability. Summary of the Invention

[0004] The present invention provides a method, device and equipment for adaptive imbalance control of the flight attitude of an unmanned aerial vehicle (UAV), aiming to realize active utilization and intelligent control of imbalance phenomena during the flight of the UAV. The method integrates key technologies such as inverse trajectory analysis, imbalance strategy generation, dynamic balance benchmark, imbalance recovery space-time manifold construction, topological feature extraction, mapping rule design, and adaptive control, and performs intelligent processing of the flight attitude imbalance phenomenon throughout the entire process. Through innovative analysis methods in the space-time dimension, imbalance identification, intelligent generation of control strategies and multi-axis coordinated linkage are realized, forming a UAV imbalance control system with predictive capability, rapid response, path optimization, effective coordination and adaptive adjustment.

[0005] A first aspect of the present invention provides a method for adaptively controlling the flight attitude of a UAV, comprising the following steps:

[0006] Acquiring the current attitude state and the target attitude state of the UAV, and performing inverse trajectory analysis on the current attitude state and the target attitude state to generate an ideal control trajectory sequence;

[0007] Obtaining an imbalance point of the ideal control trajectory sequence, determining a disturbance injection position according to the imbalance point, and generating a controllable imbalance strategy library according to the disturbance injection position;

[0008] determining an imbalance control interval according to the ideal control trajectory sequence and the controllable imbalance strategy library, generating a disturbance injection sequence according to the imbalance control interval, and establishing a dynamic balance benchmark according to the disturbance injection sequence;

[0009] Reconstructing a trajectory based on the dynamic balance reference to obtain an imbalance-recovery control trajectory, and obtaining a dynamic response feature set corresponding to the imbalance-recovery control trajectory;

[0010] Obtaining a balance recovery index corresponding to the dynamic response feature set, constructing an imbalance recovery space-time manifold based on the balance recovery index, and performing feature mapping on the imbalance recovery space-time manifold to obtain a core imbalance control feature set;

[0011] establishing an imbalance-recovery mapping rule according to the core imbalance control feature set, generating a disturbance control parameter according to the imbalance-recovery mapping rule, and establishing an adaptive mapping parameter based on the disturbance control parameter;

[0012] An inverse imbalance joint control is generated according to the adaptive mapping parameters, and a flight control signal is output according to the inverse imbalance joint control to achieve adaptive imbalance control of the flight attitude of the UAV.

[0013] A second aspect of the present invention provides a UAV flight attitude adaptive imbalance control device, comprising:

[0014] A trajectory analysis module is used to obtain the current attitude state and the target attitude state of the UAV, and perform reverse trajectory analysis on the current attitude state and the target attitude state to generate an ideal control trajectory sequence;

[0015] an imbalance acquisition module, configured to acquire an imbalance point of the ideal control trajectory sequence, determine a disturbance injection position according to the imbalance point, and generate a controllable imbalance strategy library according to the disturbance injection position;

[0016] a benchmark establishment module, configured to determine an imbalance control interval according to the ideal control trajectory sequence and the controllable imbalance strategy library, generate a disturbance injection sequence according to the imbalance control interval, and establish a dynamic balance benchmark according to the disturbance injection sequence;

[0017] a trajectory reconstruction module, configured to perform trajectory reconstruction based on the dynamic balance reference to obtain an imbalance-recovery control trajectory, and obtain a dynamic response feature set corresponding to the imbalance-recovery control trajectory;

[0018] a feature extraction module, configured to obtain a balance recovery index corresponding to the dynamic response feature set, construct an imbalance recovery spatiotemporal manifold based on the balance recovery index, and perform feature mapping on the imbalance recovery spatiotemporal manifold to obtain a core imbalance control feature set;

[0019] a mapping generation module, configured to establish an imbalance-recovery mapping rule according to the core imbalance control feature set, generate a disturbance control parameter according to the imbalance-recovery mapping rule, and establish an adaptive mapping parameter based on the disturbance control parameter;

[0020] The control execution module is used to generate an inverse imbalance joint control according to the adaptive mapping parameters, output a flight control signal according to the inverse imbalance joint control, and realize adaptive imbalance control of the flight attitude of the UAV.

[0021] The third aspect of the present invention proposes a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for adaptive imbalance control of the flight attitude of a drone disclosed in the first aspect are implemented.

[0022] The beneficial effects of the present invention are embodied in the following aspects: 1. By inferring from the target attitude state to the current attitude state through the reverse trajectory analysis technology, it is possible to accurately identify the imbalance point in the trajectory and evaluate its controllability, establish a policy library containing a variety of imbalance control modes, and transform the imbalance phenomenon of traditional passive error correction into an actively utilized control resource, thereby improving the maneuverability and control efficiency of the UAV. 2. The imbalance recovery process is mapped to a four-dimensional space-time manifold, and the geometric characteristics of the manifold are identified through topological analysis. The lengthy recovery path is compressed into an efficient control trajectory using space-time folding transformation, and the folded node set is extracted to form the core imbalance control feature, thereby improving the control efficiency and response speed, and solving the problem of lengthy recovery paths in traditional methods. 3. By establishing imbalance-recovery mapping rules and adaptive parameter fusion technology, the imbalance control strategy can be automatically adjusted according to flight status and environmental changes, and the coordinated imbalance control and unified output of the three axes of pitch, roll and yaw can be achieved, solving the problems of poor adaptability and multi-axis coordination difficulties in traditional control, and forming a complete adaptive imbalance control system.

[0023] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings herein illustrate specific examples of the technical solutions described in the present invention, and together with the specific implementation methods constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.

[0025] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.

[0026] Figure 1 The present invention is a flowchart of a method for adaptively controlling the flight attitude of a UAV.

[0027] Figure 2 This is a structural block diagram of a UAV flight attitude adaptive imbalance control device of the present invention.

[0028] Figure 3 It is a structural schematic diagram of a computer device of the present invention. DETAILED DESCRIPTION

[0029] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0030] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0031] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0032] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0033] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0034] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0035] The technical solutions of the embodiments of this application are introduced below.

[0036] like Figure 1 As shown, an embodiment of the present invention provides a method for adaptively controlling the flight attitude of a UAV, including the following steps S110 to S170:

[0037] Step S110 , obtaining the current attitude state and the target attitude state of the UAV, and performing reverse trajectory analysis on the current attitude state and the target attitude state to generate an ideal control trajectory sequence.

[0038] Specifically, the drone's current attitude state is first determined. The flight control system's inertial measurement unit (IMU) collects attitude information in real time. The IMU integrates a three-axis gyroscope, a three-axis accelerometer, and a three-axis magnetometer, with a sampling frequency of 1000Hz, ensuring high-precision updates of attitude data. The gyroscope measures the drone's angular velocity around the X, Y, and Z axes with a range of ±2000° / s and an accuracy of 0.01° / s. The accelerometer detects linear acceleration along the three axes with a range of ±16g and a resolution of 0.001g. The magnetometer measures the Earth's magnetic field strength to obtain heading angle information with an accuracy of ±0.5°. The extended Kalman filter algorithm fuses data from these three sensors to calculate the drone's real-time attitude angles, including pitch angle θ, roll angle φ, and yaw angle ψ. Angular velocity and angular acceleration information are simultaneously collected along each axis. The GPS module provides three-dimensional position coordinates with centimeter-level accuracy. The optical flow sensor assists in measuring the horizontal velocity vector, and the ultrasonic altimeter measures height above the ground. Through multi-sensor data fusion, the current attitude state is determined. During a certain flight, the current attitude state shows a pitch angle of 15.2°, a roll angle of -3.8°, and a yaw angle of 127.5°, corresponding to angular velocities of 0.8° / s, -0.2° / s, and 1.5° / s, respectively. The target attitude state of the drone is then obtained, and the desired attitude parameters are obtained through the flight mission planning module and control command analysis. The mission planning module calculates the desired attitude angle and rate of change at each moment based on the preset flight waypoints, maneuvers, and mission requirements. For cruise flight, the target attitude state generally requires the attitude angle to remain stable, with the pitch angle maintained at the horizontal flight angle, the roll angle close to zero, and the yaw angle set according to the heading requirements. For maneuvering flight, the target attitude state requires precise definition of the target value and time constraints for the attitude transition. When performing specific maneuvers, such as rapid turns, the target attitude state is set to maintain a pitch angle of 15°±2° to maintain altitude, adjust the roll angle to 25° to generate a turning moment, and change the yaw angle from the current 127.5° to the target 217.5° to complete a 90° turn. The entire attitude transition process is required to be completed within 3.5 seconds. The current and target attitude states are obtained through real-time sensor data collection and mission command analysis.

[0039] In some embodiments, the reverse trajectory analysis of the current posture state and the target posture state to generate an ideal control trajectory sequence includes: performing reverse path planning from the target posture state to the current posture state; determining intermediate control nodes based on the reverse path planning, and time-marking the intermediate control nodes; obtaining the posture change gradient of the intermediate control node; and generating an ideal control trajectory sequence based on the time-marking and the posture change gradient.

[0040] First, reverse path planning is performed from the target attitude state to the current attitude state. Using a reverse search algorithm, the target attitude state is used as the starting point and the current attitude state as the endpoint, constructing an inverse kinematics solution. Taking a turning maneuver as an example, a reverse path search is performed from the target attitude state (pitch angle 15°, roll angle 25°, yaw angle 217.5°) to the current attitude state (pitch angle 15.2°, roll angle -3.8°, yaw angle 127.5°). Compared to traditional forward planning, reverse planning has the advantage of starting from the desired final state, ensuring the accuracy and reachability of the trajectory endpoint. Reverse path planning employs an optimization search method in three-dimensional attitude space to find a feasible path from the target state to the current state while satisfying the UAV's dynamic constraints and actuator limitations. The UAV's maximum angular velocity limits (pitch velocity ≤ 30° / s, roll velocity ≤ 45° / s, yaw velocity ≤ 25° / s) and angular acceleration constraints are taken into account to ensure the physical feasibility of the planned path. A reverse search algorithm traverses the attitude space, calculating the time and energy costs of each axis angle change, and selects the path with the lowest overall cost as the optimal solution. Furthermore, reverse path planning must consider attitude coupling effects. Changes in roll angle affect yaw control efficiency, and yaw motion generates additional rolling torque. Therefore, the planning process requires a comprehensive analysis of the interplay between the three-axis attitudes. Through reverse search and constrained optimization, a reverse path plan is generated from the target attitude state to the current attitude state.

[0041] Next, intermediate control nodes are determined based on the reverse path planning and time-series labeled. Based on the three-dimensional attitude change trajectory of the reverse path planning, the intermediate control nodes are identified using an equal time interval segmentation method. The total time of 3.5 seconds in the reverse path planning is divided into seven 0.5-second time segments, and the corresponding attitude state at each time node is extracted as an intermediate control node. For example, the first intermediate control node corresponds to time t = 0.5s, with attitude angles of 15.0° pitch, 21.2° roll, and 205.1° yaw. The second node corresponds to time t = 1.0s, with attitude angles of 15.0° pitch, 17.8° roll, and 192.8° yaw, and so on until the seventh node. In addition to attitude angle information, each intermediate control node also records the corresponding angular velocity and angular acceleration values. The identified intermediate control nodes are then time-series labeled using a descending timestamp labeling method, starting from the target state and moving toward the current state. The target attitude state is marked as T0 = 3.5s, the first intermediate control node is marked as T1 = 3.0s, the second node is marked as T2 = 2.5s, and so on, until the current attitude state is marked as T7 = 0s. Timing marking not only records time information but also identifies the priority and criticality of each node. Nodes with large attitude changes are marked as critical control points, requiring precise control; nodes with gradual attitude changes are marked as transition points, allowing for a certain degree of control error. By using equal time division and decremental marking, intermediate control nodes with timing markings are obtained.

[0042] Then, the attitude change gradient of the intermediate control nodes is obtained. Based on the intermediate control nodes with time series marks, the attitude change rate and change acceleration between each node are calculated. The attitude change gradient of each axis is calculated using the attitude angle difference between adjacent intermediate control nodes and the time interval of the time series mark. The pitch angle change gradient is obtained by dividing the pitch angle difference between adjacent nodes by the time interval. From T1 to T2, the pitch angle remains unchanged at 15.0°, with a change gradient of 0° / s; the roll angle changes from 21.2° to 17.8° with a time interval of 0.5s, with a change gradient of -6.8° / s; and the yaw angle changes from 205.1° to 192.8° with a change gradient of -24.6° / s. The second-order gradient, namely the attitude change acceleration, is further calculated by the difference of the change gradient between adjacent time periods. For example, the roll angle gradient during the T0-T1 period is -7.6° / s, and during the T1-T2 period is -6.8° / s, with a second-order gradient of 1.6° / s², indicating a slowing roll angle change rate. The second-order gradient calculation of the yaw angle shows that its rate of change remains relatively stable during the maneuver. Furthermore, attitude change gradients include coupled gradient analysis, the influence coefficient of roll angle change on yaw control, and the degree of perturbation of yaw maneuvers on roll stability. Through multi-order gradient calculation and coupled analysis, attitude change gradients at intermediate control nodes are derived.

[0043] Finally, an ideal control trajectory sequence is generated based on the timing markers and attitude gradients. The complete ideal control trajectory is constructed by integrating the time constraint information of the timing markers and the dynamic characteristics of the attitude gradients. The time base of the trajectory sequence is established using the time nodes T0 to T7 in the timing markers, ensuring the precise definition of each control moment. Furthermore, the first-order and second-order derivative information of the attitude gradients is combined to design the intensity and trend of control commands for each time period. During the T1-T2 time period, the roll angle gradient of -6.8° / s and the second-order gradient of 1.6° / s² indicate that the roll control command should start at a moderate intensity and gradually decrease. During the same time period, the steady yaw angle gradient of -24.6° / s requires the yaw control command to maintain a constant intensity. The priority information in the timing markers is used to increase trajectory density at key control points, improving control accuracy. Interpolation and smoothing are used at transition points to ensure trajectory continuity. The ideal control trajectory sequence contains precise attitude target values ​​and corresponding control command amplitudes for each 0.1s moment from t=0 to t=3.5s. The trajectory sequence also integrates feedforward control information, predicting the control requirements at the next moment based on the attitude change gradient and adjusting the control output in advance. Through the time synchronization of timing markers and the dynamic matching of attitude change gradients, an ideal control trajectory sequence is generated.

[0044] Step S120 , obtaining the imbalance point of the ideal control trajectory sequence, determining the disturbance injection position according to the imbalance point, and generating a controllable imbalance strategy library according to the disturbance injection position.

[0045] Specifically, this step aims to transform the natural imbalances in the ideal control trajectory sequence into proactive control resources. By identifying key nodes in the trajectory sequence that are prone to imbalance, analyzing their imbalance characteristics and controllability, and determining the optimal timing and method for disturbance injection, we ultimately establish a strategy library encompassing multiple imbalance control modes, shifting the control approach from passive error correction to proactive imbalance exploitation.

[0046] In some embodiments, obtaining the imbalance point of the ideal control trajectory sequence includes: analyzing the posture transition key points in the ideal control trajectory sequence; determining the available natural imbalance direction at the posture transition key points; evaluating the controllability of the natural imbalance direction and generating a controllability evaluation result; and obtaining the imbalance point based on the natural imbalance direction and the controllability evaluation result.

[0047] First, based on the generated ideal control trajectory sequence, the precise attitude target values ​​at each 0.1-second moment from t=0 to t=3.5s are deeply analyzed to identify key transition points where the attitude change exceeds a preset threshold. The attitude angle change rate ω = (θ(t+Δt) - θ(t)) / Δt between adjacent time points is calculated, where θ represents the pitch angle, roll angle φ, or yaw angle ψ, and Δt=0.1s. When the pitch angle change rate |ωθ| exceeds 5° / 0.1s, the roll angle change rate |ωφ| exceeds 8° / 0.1s, or the yaw angle change rate |ωψ| exceeds 10° / 0.1s, the time point is marked as an attitude transition key point. In the ideal control trajectory sequence for a turning maneuver, key attitude transition points are identified: the roll angle begins to increase rapidly at t=0.3s, reaches a maximum of 25° at t=1.2s, the yaw angle rate of change reaches its peak at t=2.1s, the roll angle begins to decrease at t=2.8s, and the attitude approaches the target state at t=3.2s. Each key point corresponds to a significant change in the UAV's dynamic characteristics and is the moment when control deviation and imbalance are most likely to occur. In high-speed reconnaissance missions, key points for turning maneuvers are typically concentrated in the roll axis change phase, while in precision hovering missions, key points are primarily distributed during fine-tuning of various axes. Considering the specific characteristics of different maneuvers, key points for climbing maneuvers are primarily concentrated during periods of dramatic pitch axis changes, while key points for rapid evasive maneuvers manifest as simultaneous rapid changes in multiple axes. Key attitude transition points are identified through timing analysis and gradient calculation of the ideal control trajectory sequence.

[0048] Next, the available natural imbalance directions are determined at key attitude transition points. Based on the identified key attitude transition points, the UAV's dynamic characteristics and imbalance tendencies at these moments are analyzed in depth. At the roll start key point (t=0.3s), the UAV begins to establish its roll angle from a balanced state. At this point, the aircraft is most sensitive to roll axis disturbances. Natural imbalance directions are primarily manifested as overshoot or undershoot of the roll angle, and available imbalance directions are positive and negative roll axis excursions. At the roll peak key point (t=1.2s), the aircraft is at its maximum roll angle, where the coupling effect between the yaw and pitch axes is strongest. Natural imbalance directions include premature changes in the yaw angle and additional fluctuations in the pitch angle. At the yaw change key point (t=2.1s), the aircraft's yaw angular velocity reaches its maximum. At this point, the primary natural imbalance directions are yaw axis angular velocity fluctuations and secondary oscillations of the roll angle. During payload delivery missions, a noticeable tendency for pitch-axis imbalance occurs at the moment of delivery. During formation flying missions, airflow interference from adjacent drones can cause coupled imbalances in the roll and yaw axes. The manifestation of imbalance direction also differs significantly during different flight phases. Imbalance during rapid maneuvers manifests primarily as sudden changes in angular velocity, while imbalance during steady flight manifests as slow angular shifts. By analyzing the dynamic characteristics and coupling relationships of each key point, multiple primary natural imbalance directions can be identified at each transition point, yielding multiple exploitable natural imbalance directions.

[0049] Next, the controllability of the natural imbalance directions is evaluated, generating controllability assessment results. For each identified natural imbalance direction, a linearization analysis method is used to assess the controllability of each imbalance direction. The controllability assessment primarily considers three aspects: the predictability of the imbalance magnitude, the controllability of the recovery time, and the degree of impact on the overall trajectory. Taking the roll axis positive imbalance at the critical point t=0.3s as an example, simulation analysis shows that a 5° positive roll imbalance can be fully recovered within 0.8 seconds through control commands, with the impact on the subsequent trajectory lasting no more than 1.5 seconds, resulting in a high controllability rating. In contrast, the yaw axis imbalance at the critical point t=2.1s, while smaller in magnitude, requires 1.2 seconds to recover due to the high-speed yaw phase, impacting the accuracy of the second half of the turn. Therefore, the controllability rating is medium. Under complex weather conditions, such as gusty winds, the difficulty of imbalance recovery increases significantly, requiring a corresponding adjustment to the controllability rating. By establishing a three-dimensional assessment model based on imbalance amplitude A, recovery time T, and impact range R, the controllability index C is calculated as w1·f1(A) + w2·f2(T) + w3·f3(R), where f1(A) = e^(-A / A0), f2(T) = e^(-T / T0), and f3(R) = e^(-R / R0). w1, w2, and w3 are weighting coefficients, with w1+w2+w3=1. A0, T0, and R0 are normalization parameters. A controllability index is calculated for each natural imbalance direction, ranging from 0 to 1, with values ​​above 0.8 indicating high controllability, 0.5-0.8 indicating moderate controllability, and below 0.5 indicating low controllability. The assessment results show that most imbalance directions have high controllability, a few have moderate controllability, and a few have low controllability, generating a complete controllability assessment.

[0050] Finally, the imbalance point is obtained based on the natural imbalance direction and the controllability assessment results. By combining the various natural imbalance directions and the corresponding controllability assessment results, the imbalance points suitable for the control strategy are screened out. Priority is given to imbalance directions with high controllability assessment results as the main imbalance points. These imbalance points have the characteristics of controllable imbalance amplitude, short recovery time, and limited impact range. Imbalance directions with medium-level controllability can be used as alternative imbalance points under specific conditions, but additional safety margins need to be added. Imbalance directions with low controllability are not included in the imbalance point candidates for the time being to avoid control risks. For emergency obstacle avoidance scenarios, some imbalance points with medium-level controllability can be temporarily enabled to achieve rapid maneuvers. Through imbalance direction screening, multiple valid imbalance points are obtained, and each imbalance point has undergone rigorous safety and feasibility verification. For example, imbalance point FP1 is located at the roll initiation critical point at t = 0.3s, with an imbalance direction in the positive roll axis, an expected imbalance amplitude of 3-7°, and a recovery time of 0.6-1.0 seconds. Imbalance point FP2 is located at the same critical point but with an imbalance direction in the negative roll axis. Its parameters are similar but the direction is opposite. Different imbalance points also have temporal dependencies. Triggering certain imbalance points can affect the availability of subsequent imbalance points, so it is necessary to establish constraints between these imbalance points. Through natural imbalance direction analysis and controllability screening, verified imbalance points are identified.

[0051] The disturbance injection location is determined based on the imbalance point. Based on the multiple valid imbalance points obtained, the temporal position and axial characteristics of each imbalance point are analyzed to determine the optimal disturbance injection timing and method. Taking advantage of the fact that imbalance point FP1 is located at the critical roll initiation point at t=0.3s, the positive roll disturbance injection location is set 0.1 seconds before this moment, at t=0.2s. Injecting the disturbance in advance can create a synergistic effect with the upcoming natural imbalance direction, enhancing the speed and accuracy of roll angle establishment. For the negative roll imbalance characteristic of imbalance point FP2, a negative disturbance injection location is set at the same time, t=0.2s, but in the opposite direction of FP1. This is used when it is necessary to slow down the roll angle establishment speed. Considering the coupled yaw and pitch imbalances at the critical roll peak point at t=1.2s, the yaw disturbance injection location is set at t=1.1s. Taking advantage of the fact that the coupling effect is strongest when the roll angle is near its maximum value, an appropriate amount of yaw disturbance can be used to adjust the turning radius or enhance turning efficiency. In practical application scenarios, when a drone performs an emergency obstacle avoidance mission, the characteristic of the imbalance point located at the critical yaw change point at t=2.1s can be exploited to set the injection point for a rapid yaw perturbation at t=2.0s. By injecting the perturbation before the yaw angular velocity reaches its peak, a faster heading change can be achieved. In precision hovering missions, based on the distribution of imbalance points during the fine-tuning phase of each axis, the perturbation injection point is set at the beginning of attitude adjustment, helping the drone achieve the desired hovering attitude more quickly. For formation flying scenarios, when an imbalance point caused by airflow interference between adjacent drones is detected, a pre-compensation perturbation can be injected 0.3 seconds before the interference arrives to proactively offset the impact of external interference. Considering the temporal correlation between different imbalance points, adjacent perturbation injection points must be sufficiently separated by a time constraint of |t_i - t_j| ≥ Δt_min, where Δt_min = 0.4s and i ≠ j, to avoid the combined effects of multiple perturbations. By analyzing the time characteristics and axial properties of the imbalance points, the disturbance injection positions corresponding to the imbalance points are determined.

[0052] In some embodiments, generating a controllable imbalance strategy library based on the disturbance injection position includes: establishing a correspondence between the disturbance amplitude and the recovery time for each of the disturbance injection positions; determining an optimal disturbance parameter combination based on the correspondence; and classifying and storing the optimal disturbance parameter combination to generate a controllable imbalance strategy library.

[0053] First, a relationship between disturbance amplitude and recovery time was established for each disturbance injection location. Based on the determined disturbance injection locations, the effects of different disturbance amplitudes on the control process were analyzed in depth. For the roll axis forward disturbance injection location at t=0.2s, theoretical calculations and experimental verification revealed that small disturbances can quickly recover to the ideal trajectory but have relatively limited control effects. Medium-amplitude disturbances strike a good balance between control effect and recovery time. Large disturbances, while producing significant control effects, require longer recovery times. For the yaw axis disturbance injection location at t=1.1s, due to the peak roll phase and significant multi-axis coupling effects, disturbances of the same amplitude produce significantly longer recovery times than other injection locations. Therefore, more caution is required in the selection of disturbance amplitudes. In search and rescue missions, the fast yaw disturbance injection location at t=2.0s exhibits unique advantages. By leveraging the dynamic characteristics of the yaw angular velocity near its peak, a smaller disturbance amplitude can produce a significant directional change while maintaining a relatively short recovery time. Different application scenarios require different balances between disturbance amplitude and recovery time. Emergency obstacle avoidance tasks can tolerate longer recovery times to achieve greater maneuverability, while precise positioning tasks require shorter recovery times to maintain control accuracy. Through systematic parameter analysis and multi-scenario verification, a correlation between disturbance amplitude and recovery time covering different amplitude levels is established for each disturbance injection location.

[0054] Then, the optimal perturbation parameter combination was determined based on the corresponding relationship. Based on the established relationship between perturbation amplitude and recovery time, the optimal parameter configuration for each injection location was screened, taking into account control effectiveness, safety requirements, and mission characteristics. For the roll axis perturbation injection location at t=0.2s, the medium-amplitude perturbation was selected as the optimal parameter based on the analysis results in the corresponding relationship: small perturbations quickly recover but have limited control effectiveness, medium-amplitude perturbations have better balance, and large-amplitude perturbations have significant control effectiveness but long recovery times. This parameter not only produces a significant roll angle establishment effect but also ensures complete recovery to the desired trajectory within a reasonable time. For the yaw axis perturbation injection location at t=1.1s, the small-amplitude perturbation was selected as the optimal parameter, considering that the recovery time at this location is significantly longer than that at other locations in the corresponding relationship. This parameter achieves sufficient control gain through the coupling effect while avoiding the impact of excessive recovery time on subsequent control nodes. For the fast yaw perturbation injection location at t=2.0s, the small-amplitude perturbation was selected as the optimal parameter based on the analysis in the corresponding relationship that a small perturbation amplitude can produce a significant effect with a relatively short recovery time. In formation flying scenarios, this configuration enables rapid heading adjustments without compromising the overall stability of the formation. The optimization process must consider the temporal relationships between multiple disturbance injection locations. When multiple locations are used consecutively, parameter configurations must be appropriately adjusted based on the analysis of their respective relationships to avoid degradation in control performance due to cumulative effects. Multi-objective optimization techniques are used to determine the optimal combination of disturbance parameters.

[0055] Finally, the optimal disturbance parameter combinations are categorized and stored to form a library of controllable imbalance strategies. Based on the optimal disturbance parameter combinations determined for each disturbance injection location, these are systematically categorized and organized according to disturbance axial characteristics, application scenario type, and control objective requirements. A roll-axis imbalance strategy classification is established, integrating medium-amplitude positive and negative disturbance parameter combinations based on the t=0.2s position. These are specifically designed for various flight maneuvers that require rapid establishment or suppression of the roll angle, such as rapid turning maneuvers, lateral avoidance maneuvers, and precise attitude adjustments. A yaw-axis imbalance strategy classification is created, uniformly managing small-amplitude coupled disturbance parameter combinations at the t=1.1s position and small-amplitude rapid disturbance parameter combinations at the t=2.0s position, respectively addressing the different control requirements of precise turning radius adjustment and rapid heading changes. A multi-axis coordinated imbalance strategy classification is established, integrating multiple optimal disturbance parameter combinations according to reasonable timing relationships and axial coordination principles to form a comprehensive control scheme suitable for complex maneuvering tasks. For example, in dynamic target tracking missions, collaborative strategies can simultaneously utilize the combined effects of medium-amplitude perturbations in the roll axis and small-amplitude perturbations in the yaw axis to achieve fast and precise tracking maneuvers. Each strategy classification contains complete parameter configuration information, detailed descriptions of applicable conditions, clear safety restrictions, and accurate performance expectations. The strategy library establishes a multi-dimensional intelligent indexing mechanism to support rapid retrieval of the most suitable imbalance control strategy based on the current flight status, specific mission type, environmental condition changes, and immediate control objectives. Through the scientific classification and systematic storage of optimal perturbation parameter combinations, a controllable imbalance strategy library covering a variety of application scenarios is formed.

[0056] Step S130 : determining an imbalance control interval according to the ideal control trajectory sequence and the controllable imbalance strategy library, generating a disturbance injection sequence according to the imbalance control interval, and establishing a dynamic balance benchmark according to the disturbance injection sequence.

[0057] Specifically, the imbalance control interval is determined based on the ideal control trajectory sequence and the controllable imbalance strategy library. Based on the precise attitude target values ​​at every 0.1s from t=0 to t=3.5s in the generated ideal control trajectory sequence, combined with the various imbalance strategies in the controllable imbalance strategy library, the time periods in the trajectory sequence suitable for applying imbalance control are analyzed. By analyzing the attitude change characteristics of the ideal control trajectory sequence, intervals with large attitude adjustments and long durations are identified as candidate time periods for imbalance control. In the interval from t=0.2s to t=1.5s, the ideal control trajectory sequence shows a rapid increase in the roll angle from the initial state to a peak of 25°. The sustained attitude change characteristics in this interval fully match the application conditions for the roll axis imbalance strategy classification in the controllable imbalance strategy library. Using the medium-amplitude positive perturbation parameter combination of the roll axis imbalance strategy in the strategy library, this time period is identified as the roll axis-dominated imbalance control interval. In the interval t=1.8s to t=3.2s, the ideal control trajectory sequence exhibits a rapid yaw angle transition from 127.5° to 217.5°, accompanied by a gradual roll angle reduction. This complex multi-axis attitude transition characteristic corresponds precisely to the design objectives of the multi-axis coordinated imbalance strategy classification in the controllable imbalance strategy library. Combining the small, rapid perturbation parameter combination of the yaw axis imbalance strategy and the negative perturbation parameter combination of the roll axis strategy in the strategy library, this interval is identified as the multi-axis coordinated imbalance control interval. In target tracking tasks, when the ideal control trajectory sequence indicates the need for frequent, small attitude adjustments, the precise control strategy category in the strategy library can be used to set the corresponding imbalance control interval. For the initial acceleration phase from t=0 to t=0.2s and the final stabilization phase from t=3.2s to t=3.5s in the ideal control trajectory sequence, since the attitude changes are relatively gradual and there are no corresponding imbalance strategies in the strategy library, these periods are excluded from the imbalance control interval, maintaining the conventional control mode. The imbalance control range is determined by analyzing the time characteristics of the ideal control trajectory sequence and verifying the strategy matching of the controllable imbalance strategy library.

[0058] A disturbance injection sequence is generated based on the imbalance control interval. Based on the determined imbalance control interval, the imbalance strategy within each interval is converted into specific time-sequenced disturbance commands. For the dominant roll axis imbalance control interval from t=0.2s to t=1.5s, a continuous disturbance injection command is designed within this interval using the corresponding medium-amplitude positive disturbance parameter combination of the roll axis imbalance strategy. First, the primary roll axis positive disturbance is injected at t=0.2s. The disturbance signal d(t) = δ*·u(t-t0)·[1-u(t-t0-T_d)], where δ* is the optimal disturbance amplitude, u(t) is a unit step function, t0=0.2s is the injection time, and T_d is the duration. The disturbance amplitude is set to a medium level, covering the critical stage of roll angle establishment. Auxiliary roll disturbances are then injected at t=0.8s and t=1.2s, with gradually decreasing amplitudes, to maintain and optimize the roll angle establishment process. For the multi-axis coordinated imbalance control interval from t=1.8s to t=3.2s, a multi-axis joint disturbance injection command is designed based on the combination of small-amplitude rapid disturbances of the yaw axis and negative disturbance parameters of the roll axis. At t=1.8s, a small-amplitude disturbance of the yaw axis and a negative disturbance of the roll axis are simultaneously started to achieve the multi-axis coordinated imbalance effect; at t=2.3s, an enhanced disturbance of the yaw axis is injected to accelerate the change process of the yaw angle; at t=2.8s, a final disturbance of the roll axis is provided to assist in the smooth reduction of the roll angle. In forest fire monitoring tasks, when the observation angle needs to be adjusted quickly, additional pitch axis disturbance injection commands can be added to the multi-axis coordinated imbalance control interval to form a three-axis coordinated disturbance sequence with the existing yaw and roll disturbances. For the periods t=0 to t=0.2s and t=3.2s to t=3.5s, which are excluded from the imbalance control interval, the disturbance injection sequence does not contain any imbalance-related instructions, and the instruction output of the conventional control mode is maintained. The disturbance injection sequence also takes into account the connection relationship between the various imbalance control intervals, and sets a smooth transition of the disturbance intensity in the transition period from t=1.5s to t=1.8s to avoid control mutations caused by interval switching. Each disturbance injection instruction contains a clear time stamp, axial designation, amplitude setting, and duration, forming a complete set of timing control instructions. Through strategy conversion and timing processing of the imbalance control interval, a disturbance injection sequence covering the entire control process is generated.

[0059] In some embodiments, establishing a dynamic balance benchmark based on the disturbance injection sequence includes: determining an expected recovery trajectory after imbalance based on the disturbance injection sequence; generating a tolerance range based on the expected recovery trajectory; determining a time reference point based on the expected recovery trajectory; and combining the expected recovery trajectory, the tolerance range, and the time reference point to form a dynamic balance benchmark.

[0060] First, the expected recovery trajectory after the imbalance is determined based on the disturbance injection sequence. For a positive roll axis disturbance injection command at t = 0.2 s, dynamic modeling and trajectory prediction determine that this medium-amplitude disturbance causes the roll angle to deviate from the ideal trajectory by approximately 4-6° within 0.3 s. It then gradually recovers to the desired state within 0.9 s, following the exponential decay law θ(t) = θ_ref + (θ_max - θ_ref)·e^(-t / τ), where θ_ref is the reference trajectory value, θ_max is the maximum deviation value, and τ is the time constant. The entire process exhibits a trajectory characteristic of rapid deviation followed by smooth convergence. For a multi-axis coordinated disturbance injection command at t = 1.8 s, the expected recovery trajectory exhibits complex multi-axis coupled evolution due to the simultaneous involvement of a small yaw axis disturbance and a negative roll axis disturbance. The yaw angle recovery process affects the convergence speed and trajectory shape of the roll angle through aerodynamic coupling effects, extending the overall recovery time to 1.2 s. In target tracking tasks, when rapid adjustments to the observation angle are required, the accuracy of the predicted recovery trajectory directly impacts the continuity of the tracking effect. By analyzing the intensity, duration, and interaction of disturbances at each moment in the disturbance injection sequence, and considering the superposition and timing effects of adjacent disturbances, we establish the attitude offset, recovery time constant, and trajectory convergence characteristics corresponding to each disturbance, forming an expected recovery trajectory that covers the entire control process.

[0061] Then, a tolerance range is generated based on the expected recovery trajectory, and a time reference point is determined based on the expected recovery trajectory. Using the determined expected recovery trajectory, the uncertainty factors and random disturbance effects in the trajectory recovery process are deeply analyzed, and a scientific and reasonable allowable deviation range is set. For the expected recovery trajectory of the roll axis disturbance at t=0.2s, taking into account practical factors such as atmospheric turbulence, sensor noise, and actuator response delay, the tolerance range of the roll angle recovery process is set to ±2°, and the tolerance range of the yaw angle and pitch angle is set to ±1°, ensuring that the system can still maintain stable control under normal disturbance conditions. For the expected recovery trajectory of the multi-axis coordinated disturbance at t=1.8s, due to the complex multi-axis coupling effect and relatively long recovery time, the yaw angle tolerance range is expanded to ±3°, and the roll angle tolerance range is set to ±2.5°, providing ample adjustment space for the multi-axis coordinated recovery process.

[0062] Next, based on the temporal evolution characteristics of each expected recovery trajectory, the time reference points are defined as follows: t_peak = argmax|θ(t) - θ_ref(t)| (deviation reference point), t_half = {t | |θ(t) - θ_ref(t)| = 0.5·|θ_max - θ_ref|} (half-way reference point), and t_recover = {t | |θ(t) - θ_ref(t)| < ε_final} (recovery reference point). Key time nodes in the trajectory recovery process are identified as monitoring and evaluation reference points. For the expected recovery trajectory of the roll axis perturbation, the maximum deviation at 0.3 seconds is set as the deviation reference point, the full recovery at 0.9 seconds is set as the recovery reference point, and the halfway return at 0.6 seconds is set as the transition reference point. For the expected recovery trajectory of a multi-axis coordinated perturbation, the multi-axis peak deviation at 0.4 seconds is set as the composite deviation reference point, the dominant axis recovery at 0.8 seconds is set as the mid-term reference point, and the full recovery at 1.2 seconds is set as the final reference point. In formation flying scenarios, time reference points must be synchronized with the formation coordination rhythm to ensure that individual aircraft recovery does not affect formation stability. Through detailed time analysis of the expected recovery trajectory, key time reference points are determined.

[0063] Finally, the expected recovery trajectory, tolerance range, and time reference point are combined to form a dynamic balance benchmark. A comprehensive dynamic balance evaluation standard system is established by comprehensively determining each expected recovery trajectory, corresponding tolerance range, and key time reference point. The expected recovery trajectory for a t=0.2s disturbance, a ±2° roll angle tolerance range, a deviation reference point (0.3 seconds), and a recovery reference point (0.9 seconds) are organically combined to form a dynamic balance benchmark for roll axis imbalance, providing a full-process monitoring standard for roll axis imbalance control. The expected recovery trajectory for a t=1.8s multi-axis disturbance, a ±3° yaw angle and ±2.5° roll angle tolerance range, a composite deviation reference point (0.4 seconds), and a final reference point (1.2 seconds) are systematically combined to form a dynamic balance benchmark for multi-axis coordinated imbalance, providing a comprehensive evaluation basis for complex multi-axis imbalance control. The dynamic balance benchmark is different from the traditional static balance standard. It innovatively allows for expected deviations in attitude within a preset controllable range, and provides clear recovery path guidance and time node requirements, realizing a shift in control thinking from passive error correction to active utilization of imbalance. In the actual control process, when the UAV attitude deviates beyond the tolerance range of the dynamic balance benchmark or the recovery time exceeds the benchmark point requirements, the control decision mechanism will automatically trigger the corresponding adjustment strategy or initiate safety protection measures. In complex environments such as search and rescue missions, the dynamic balance benchmark can adapt to rapidly changing flight needs and maximize maneuverability while ensuring safety. Through the organic combination of expected recovery trajectory, tolerance range and time benchmark point, a dynamic balance benchmark that adapts to the characteristics of imbalance control is established.

[0064] Step S140 : performing trajectory reconstruction based on the dynamic balance reference to obtain an imbalance-recovery control trajectory, and obtaining a dynamic response feature set corresponding to the imbalance-recovery control trajectory.

[0065] Specifically, the dynamic balance benchmark is converted into an executable control trajectory. Trajectory reconstruction technology transforms the ideal linear control process into a nonlinear control process that includes active imbalance and precise recovery. This reconstructed trajectory fully exploits the imbalance phenomenon to enhance the UAV's maneuverability while maintaining the predictability and safety of the control process.

[0066] In some embodiments, the trajectory reconstruction based on the dynamic balance benchmark to obtain the imbalance-recovery control trajectory includes: performing trajectory inversion in the imbalance stage according to the dynamic balance benchmark to generate an inversion trajectory segment; determining the imbalance-recovery transition node based on the inversion trajectory segment; and performing trajectory splicing according to the transition node and the dynamic balance benchmark to generate the imbalance-recovery control trajectory.

[0067] The trajectory of the imbalance phase is back-calculated based on the dynamic balance benchmark, inferring the imbalance development process from the desired recovery endpoint. The back-calculation algorithm is based on inverse time dynamics ẋ=-f(x,u), starting from the equilibrium state x_balance and working backwards to the initial imbalance state x_imbalance. The back-calculation process considers system controllability and reachability constraints to ensure that the generated trajectory is physically feasible. Constraints include attitude angle constraints |φ|≤30° and |θ|≤30°, angular velocity constraints |ω|≤100° / s, and actuator saturation limits. Boundary conditions are set within the tolerance range of the dynamic balance benchmark to ensure consistency between the back-calculation starting point and the expected recovery trajectory. The numerical solution uses the fourth-order Runge-Kutta method with a step size of 0.001 second to ensure accuracy. The time span of the back-calculated trajectory segment is adaptively adjusted based on the imbalance severity: 1-2 seconds for mild imbalance, 2-3 seconds for moderate imbalance, and 3-5 seconds for severe imbalance. The trajectory is verified to be reasonable through energy conservation and Lyapunov stability checks. For example, in a crosswind disturbance scenario during agricultural plant protection operations, the complete evolution of an initial 15° roll imbalance state can be obtained by back-calculating 2.5 seconds from the target equilibrium state. Through inverse dynamics calculations and multi-constraint verification, a back-calculated trajectory segment accurately reflects the development of the imbalance.

[0068] Based on back-calculated trajectory segment analysis, the transition nodes between imbalance and recovery are determined, identifying the optimal timing for control strategy switching. Transition node identification utilizes a combination of phase plane analysis and energy analysis. On the attitude angle-angular velocity phase plane, the maximum value of the trajectory curvature κ(t) = |ẋ×ẍ| / |ẋ|³ is sought. This point typically corresponds to the critical position for changing control direction. The energy criterion calculates the total system energy E(t) = T(t) + V(t), where T is kinetic energy and V is potential energy. Candidate transition points are identified when dE / dt changes from positive to negative. Control input analysis detects sign changes and amplitude abrupt changes in u(t). A transition is confirmed when sign(u(t)) ≠ sign(u(t-Δt)) and |Δu| > u_threshold. Stability verification calculates the derivative of the Lyapunov function V(x) at the candidate point to ensure that dV / dt < 0, ensuring stability after switching. The optimal transition node is determined by fusion of multiple indicators using a weighted score (S = w1·κ_score + w2·E_score + w3·u_score), with weights w1 = 0.4, w2 = 0.3, and w3 = 0.3 reflecting the importance of each indicator. Precise positioning of the transition node is achieved using a golden section search, iterating within an initial estimate ±0.1 seconds with an accuracy of 0.01 seconds. For example, a typical pitch imbalance case determines the transition node at t = 1.35 seconds, at which point the pitch angle reaches its maximum deviation of 12°, the angular velocity approaches zero, and the control torque is poised for reversal. Through multi-dimensional analysis and precise search, a dynamically optimized imbalance-recovery transition node is determined.

[0069] Trajectory splicing is performed based on the determined transition nodes and dynamic balance benchmark to generate a smooth and continuous complete control trajectory. Trajectory splicing leverages three core elements of the dynamic balance benchmark: the expected recovery trajectory is used as the splicing target reference to ensure the overall trend of the generated trajectory is consistent with the expected trend; a tolerance range (±2° for roll, ±3° for yaw, and ±1° for pitch) is used as the splicing constraint boundary; and a time reference point (0.3 seconds from the reference point and 0.9 seconds from the reference point) is used as the key splicing anchor point. The splicing uses cubic spline interpolation to ensure C² continuity, strictly matching position, velocity, and acceleration constraints at the transition nodes. The spline parameters are determined by solving the linear equation system Ax=b, where A is a tridiagonal matrix, x is the coefficient to be determined, and b is the boundary condition vector extracted from the expected recovery trajectory based on the dynamic balance benchmark. The transition zone is designed using a sigmoid weight function w(t) = 1 / (1 + exp(-(t-t_switch) / τ)), where t_switch is taken from the time reference point of the dynamic balance benchmark, and the time constant τ = 0.1 second controls transition smoothness. Splicing verification directly uses the tolerance range of the dynamic balance benchmark for continuity checks: |x(t⁺)-x(t⁻)| < 2° (roll) or 3° (yaw), |ẋ(t⁺)-ẋ(t⁻)| < 0.01 rad / s², |ẍ(t⁺)-ẍ(t⁻)| < 0.1 rad / s². Dynamic consistency is verified by substituting the spliced ​​trajectory into the original dynamic equations. The residual ‖ẋ-f(x,u)‖ < δ_dyn confirms physical feasibility. Constraint satisfaction checks ensure that the entire trajectory remains within the tolerance range of the dynamic balance benchmark. Performance optimization involves fine-tuning the stitching parameters to minimize the deviation of the trajectory from the expected recovery trajectory at the time reference point. The complete trajectory output includes the time series t∈[0,T_final], the state trajectory x(t), the control trajectory u(t), and the performance evolution J(t). For example, in a building-circling mission for urban delivery, precise stitching generated a smooth 3.8-second trajectory from an initial 20° roll imbalance to full recovery, accurately passing through the preset state at the 0.3-second and 0.9-second time reference points. Through precise stitching and comprehensive verification based on a dynamic balance benchmark, a smooth and feasible imbalance-recovery control trajectory was generated.

[0070] Obtain the dynamic response feature set corresponding to the imbalance-recovery control trajectory. Based on the generated imbalance-recovery control trajectory, extract the dynamic response characteristics and behavior patterns of the UAV during the trajectory execution. For the roll axis imbalance-recovery control trajectory including the entire process of active imbalance and controllable recovery, starting from the disturbance injection moment t=0.2s, calculate the response parameters of the roll angle in the imbalance stage, and obtain the rise time of the roll angle response of 0.25 seconds, the maximum deviation of 6.2°, and the instantaneous error of 4.8° at the 0.3-second transition node. Identify the convergence law of the recovery stage from the 0.3-second transition node to the 0.9-second recovery reference point, and fit the second-order system response θ(t) = θss + (θ0 - θ ss )e^(-ζω n t)[cos(ω a t) + (ζ / √(1-ζ²))sin(ω a t)], where ω n is the natural frequency, ζ is the damping ratio, ω a = ω n √(1-ζ²), determining the dominant time constant of the recovery process τ = 1 / (ζω n ) = 0.35 seconds, and the damping ratio ζ = 0.72, showing a slightly underdamped convergence characteristic. For the multi-axis coordinated imbalance-recovery control trajectory, the coupling response phenomenon from the multi-axis disturbance injection at t = 1.8s to the complete recovery process was studied. It was found that the response delay of the yaw angle was about 0.15 seconds behind the roll angle, and the cross-coupling coefficient between the two axes was 0.31, reaching the maximum coupling strength at the 0.4-second composite deviation conversion node. The frequency response characteristics of the imbalance-recovery control trajectory were extracted through frequency domain transformation technology. The main frequency component of the roll axis imbalance process was concentrated around 2.1Hz, while the frequency component of the recovery process shifted to a low frequency of 1.3Hz. In the case of multi-axis coordination, the frequency response of the yaw and roll axes showed a 0.4Hz difference frequency coupling phenomenon. In forest fire monitoring tasks, this frequency characteristic can maintain the stability of image acquisition while quickly adjusting the observation angle, avoiding image blurring caused by high-frequency oscillation. The amplitude variation characteristics of the imbalance-recovery control trajectory were quantified. The maximum amplitude response of the roll axis imbalance was 6.2°, with an amplitude decay rate of -8.3° per second. The amplitude variation rate during the recovery process exhibited an exponential decay pattern, ensuring the predictability and safety of the maneuver. In formation flying scenarios, the consistency of the dynamic response characteristics helped maintain formation stability, with the response delay differences between aircraft kept within 0.1 seconds. The dynamic response characteristics of the imbalance-recovery control trajectory were obtained through time domain calculation, frequency domain transformation, and amplitude quantification.

[0071] Step S150 , obtaining a balance recovery index corresponding to the dynamic response feature set, constructing an imbalance recovery spatiotemporal manifold based on the balance recovery index, and performing feature mapping based on the imbalance recovery spatiotemporal manifold to obtain a core imbalance control feature set.

[0072] Specifically, based on a dynamic response feature set, key performance indicators reflecting the drone's recovery from an imbalanced state to a balanced state are extracted. This dynamic response feature set includes multidimensional data such as the drone's time response, frequency response, amplitude variation, and multi-axis coupling characteristics during the imbalance-recovery process. Balance recovery index extraction utilizes feature transformation and induction methods to extract key indicators for evaluating recovery performance from existing dynamic features. Macro-time indicators are directly extracted from the time response features: the time constant τ is used as a benchmark indicator of recovery speed, with the total recovery time T_recovery = nτ, where n is the engineering convergence coefficient, typically 3-5; the rise time t_r is directly used as a response speed indicator; and the convergence rate λ_conv = |k_decay| / A_max is derived based on the amplitude decay rate k_decay, where A_max is the maximum deviation amplitude. Meso-level dynamic indicators are converted from frequency and damping characteristics: the damping ratio ζ is directly used as the oscillation suppression indicator; the recovery frequency f_recovery is used as the system oscillation frequency; the relative overshoot M_r = A_max / |A_max-e_instant| is calculated based on the maximum deviation A_max and the instantaneous error e_instant, reflecting the degree of overshoot in the dynamic response. Micro-level coupling indicators are extracted from multi-axis characteristics: the cross-coupling coefficient K_couple is used as an indicator of the strength of inter-axis influence; the coupling complexity C_complex = t_delay × f_diff is defined based on the yaw response delay t_delay and the difference frequency coupling f_diff to quantify the difficulty of multi-axis coordination; and the ratio of the imbalance frequency f_imbalance to the recovery frequency f_recovery, R_freq = f_imbalance / f_recovery, is used as the control mode transition characteristic. For example, during agricultural plant protection operations under strong wind disturbances, indicators extracted from dynamic response characteristics showed that a recovery time constant of 0.35 seconds corresponds to rapid response, a damping ratio of 0.72 indicates good oscillation suppression, and a frequency ratio of 1.62 reflects clear control mode switching. By directly utilizing dynamic response characteristics, mathematically transforming them, and combining them for derivation, we obtain a balance recovery indicator that comprehensively reflects the performance of the imbalance recovery process.

[0073] Based on the extracted balance recovery index, the imbalance recovery space-time manifold is constructed to map the discrete recovery process data to a continuous geometric space. The space-time manifold is constructed with the balance recovery index as the coordinate dimension, and the four most representative indicators are selected to form a four-dimensional manifold space M⊂R 4The first dimension is the time constant τ, reflecting the speed of recovery; the second dimension is the damping ratio ζ, characterizing the oscillation characteristics; the third dimension is the coupling coefficient K_couple, characterizing multi-axis correlations; and the fourth dimension is the normalized time t / T_recovery, unifying the time scale. Each point p(τ,ζ,K_couple,t / T_recovery) on the manifold represents the comprehensive performance state at a certain moment in the recovery process. The imbalance recovery trajectory forms a continuous curve γ(s) on the manifold from the initial imbalance state p_imbalance to the final equilibrium state p_balance, where s is the arc length parameter. The manifold metric tensor g_μν is designed based on the physical meaning and dimensional differences of each indicator, using a weighted Euclidean metric to ensure comparability between different indicators. The manifold curvature reflects the complexity of the recovery process by calculating the curvature of the trajectory in the indicator space. Regions of high curvature correspond to critical stages where performance indicators experience drastic changes. The dynamic structure of the manifold is constructed based on the convergence rate λ_conv and the oscillation frequency f_recovery. A vector field V = λ_conv∂ / ∂τ + 2πf_recovery∂ / ∂ζ ​​is defined to describe the indicator's evolutionary direction. The manifold's topology is determined by analyzing the distribution of different recovery modes in the indicator space. Rapid response mode, stable coordination mode, and adaptive adjustment mode form distinct characteristic regions on the manifold. The manifold boundary is determined by the physical constraints of the indicator, such as the damping ratio 0 < ζ < 1 to ensure stability and the coupling coefficient |K_couple| < 1 to ensure controllability. Through geometric representation and topological analysis of the balance recovery indicator, a space-time manifold is constructed that reflects the essential characteristics of the imbalance recovery process.

[0074] In some embodiments, the feature mapping based on the imbalance recovery space-time manifold to obtain the core imbalance control feature set includes: performing a topological analysis on the imbalance recovery space-time manifold to obtain manifold geometric features; performing a space-time folding transformation based on the manifold geometric features to generate a compressed recovery path; performing key point extraction on the compressed recovery path to obtain a folded node set; and performing feature mapping based on the folded node set to obtain the core imbalance control feature set.

[0075] Topological analysis of the imbalance-recovery spacetime manifold reveals its geometric characteristics. This topological analysis employs differential geometry methods to systematically study the intrinsic properties and extrinsic shape of the manifold, extracting geometric invariants that reflect the essential laws governing imbalance recovery. Curvature analysis characterizes the curvature of the manifold by calculating the Gaussian curvature K=det(g^{-1}R) / 2 and the mean curvature H=tr(g^{-1}h) / 2, where g is the metric tensor, R is the Ricci tensor, and h is the second fundamental form. Regions where the Gaussian curvature K>0 correspond to stable recovery segments, those where K<0 represent unstable transition segments, and those where K=0 represent uniform recovery. The geodesic curvature κ_g=|dT / ds| quantifies the degree to which the recovery trajectory deviates from the geodesic line; smaller κ_g indicates more optimized control. Homology group calculation H_k(M) identifies k-dimensional "hole" structures in the manifold, where H_0 reflects the number of connected components, H_1 corresponds to the number of loops, and H_2 represents the number of cavities. For example, the imbalance recovery manifold analysis of drones performing express delivery between urban buildings caused by the complex airflow between buildings shows that there are three high curvature regions (K>2.5) corresponding to sharp turning points, five saddle points (K<-1.5) that identify the control mode switching positions, and the first Betti number b_1=2, indicating the existence of two independent recovery loops. Critical point analysis identifies extreme points, saddle points, and inflection points on the manifold by calculating the gradient ∇f=0, where f is a scalar function defined on the manifold, such as recovery time or energy consumption. The Morse index calculates the number of negative eigenvalues ​​of the Hessian matrix at the critical point, classifies the critical point type, and predicts its stability. The Riemann volume of the manifold V_M=∫_M√det(g)d 4 x quantifies the "size" of the state space; smaller volumes indicate a more concentrated and efficient recovery process. The principal curvature direction is determined by solving the characteristic equation det(h - κg) = 0, indicating the primary evolutionary direction of the recovery process. Through topological and geometric analysis of the system, the geometric characteristics of the manifold are obtained, which fully describe its structural features.

[0076] A compressed recovery path is generated by performing a spacetime folding transformation based on the geometric characteristics of the manifold. The spacetime folding transformation leverages the geometric characteristics of the manifold to compress the original, lengthy recovery path into a shorter, equivalent path through mathematical transformation. The folding transformation is defined as a diffeomorphic mapping Φ:M→M', which preserves the topological properties but modifies the metric structure such that d_M'(Φ(p),Φ(q))≤d_M(p,q) holds for all point pairs. Folding point selection is based on the curvature extremum criterion. When |K(p)|>K_threshold=2.0, point p becomes a folding candidate point, and the folding direction and angle are determined through local analysis. The folding operation is implemented using the exponential mapping exp_p:T_pM→M, which maps a vector v in the tangent space to a point exp_p(v) on the manifold. The folding path is γ_fold(t)=exp_p(tv), where t∈[0,1]. Compression of the time dimension is achieved through the Lorentz transformation: τ' = γ_L(τ - vx / c²), where γ_L = 1 / √(1 - v² / c²) is the Lorentz factor and v is the characteristic velocity, achieving "contraction" of the time dimension. Folding of the spatial dimension uses the Ricci flow equation ∂g_ij / ∂t = -2R_ij to drive metric evolution, shrinking the manifold toward a more regular shape. For example, in a forest fire monitoring task, the original recovery path length was 8.5 units, which was compressed to 3.2 units after three folding transformations, a compression ratio of 62%. The first folding utilized the high curvature (K = 3.8) of the pitch-roll coupling region to shorten the path by 35%. Multiple foldings are achieved by recursively applying transformations, and the compression effect and stability are evaluated after each folding. Through systematic folding transformations and path optimization, a compressed recovery path with space-time compression is generated.

[0077] Keypoint extraction is performed on the compression recovery path to obtain a set of collapsed nodes. Keypoint extraction employs an identification method that combines geometric features with dynamic significance to select the core nodes that determine the recovery process from the compression path. Geometric keypoints are identified through curvature analysis, including curvature maxima (κ>κ_max=2.5) to mark sharp turns, curvature zeros (|κ|<0.1) to indicate straight line segments, and curvature sign changes to mark turn transitions. Dynamic keypoints are determined based on phase space analysis and include locations with special dynamic significance, such as fixed points, limit cycles, and bifurcation points. Collapsed intersections, where multiple paths converge, are of great significance and are identified by calculating the inter-path distance d(γ_i(t),γ_j(t))<ε_merge=0.2. Topological keypoints include representative points of homotopy equivalence classes, corresponding points of generators of fundamental groups, and branch points in the covering space. For example, in an offshore wind turbine inspection, the complex imbalance recovery process caused by strong sea breezes was compressed and 12 key points were extracted: four geometric turning points (κ > 3.0), three mode switching points (control law change locations), two folding centers (multi-path convergence), and three energy extreme points (local control energy minima). The importance of these key points was calculated using a multi-index weighted method: I_p = w_1·κ_norm + w_2·E_impact + w_3·T_influence, where κ_norm is the normalized curvature, E_impact is the energy impact factor, and T_influence is the temporal influence range. The weights w_1 = 0.4, w_2 = 0.3, and w_3 = 0.3. The connectivity between nodes is represented by the reachability matrix A_ij, where A_ij = 1 indicates that node j is directly reachable from node i, forming a folded node topology network. A node's temporal attributes record its original timeline position t_original and its compressed position t_compressed. The compression ratio r_i = t_compressed / t_original reflects the degree of temporal folding. A node's state attributes contain the complete pose information, control inputs, constraints, and other information for that point, ensuring the integrity of node information. Through multi-dimensional analysis and importance assessment, the set of folded nodes that form the core structure of the compression path is extracted.

[0078] Based on the collapsed node set, feature mapping is performed to obtain the core imbalance control feature set. Feature mapping uses nonlinear dimensionality reduction and pattern recognition methods to project the high-dimensional node attribute space into the low-dimensional feature space to extract the most representative control features. The mapping function is designed as F:R n →R ᵐ(m << n), implemented by Kernel Principal Component Analysis (KPCA), with the kernel function selected as the Gaussian radial basis k(x_i, x_j) = exp(-||x_i - x_j||² / 2σ²), and the bandwidth parameter σ optimized through cross - validation to obtain σ_opt = 1.5. The first step of feature extraction is to calculate the kernel matrix K_ij = k(n_i, n_j), where n_i is the attribute vector of the i - th folding node, and then solve the eigen - equation Kα = λα to extract the principal eigen - vectors. The proportion of the explained variance of the first m principal components reaches more than 95%. Usually, m = 5 - 8 can capture the main features. The extracted core features include: the imbalance mode feature f_mode identifies typical imbalance types (uniaxial / multi - axial / burst / progressive) through clustering analysis; the recovery strategy feature f_strategy characterizes the controller selection mode (PID / adaptive / robust / intelligent); the time - scale feature f_time reflects the separation of fast and slow dynamics (fast variables < 0.5s / medium - speed variables 0.5 - 2s / slow variables > 2s); the energy distribution feature f_energy describes the distribution ratio of control energy in each channel; the stability margin feature f_margin quantifies the safety margin of the system from the instability boundary. For example, in the fine operation of power line inspection, the extracted core features show that the dominant imbalance mode is wind - induced roll (accounting for 45%), the optimal recovery strategy is adaptive sliding - mode control (success rate 92%), the key time - scale is 0.8 seconds (fast response requirement), the energy proportion of the roll channel is 68% (main control channel), and the stability margin is 15° (safety boundary). The physical meaning of the features is verified through inverse mapping to ensure that the extracted features can reconstruct the key characteristics of the original control process. The discriminability of the features is evaluated by the Fisher discriminant ratio J=(μ_1 - μ_2)² / (σ_1² + σ_2²). J > 2 indicates that the feature has good classification ability. Through the feature extraction and verification analysis of the system, a core imbalance control feature set that highly condenses the essential laws of imbalance control is obtained.

[0079] Step S160, establish an imbalance - recovery mapping rule according to the core imbalance control feature set, generate perturbation control parameters according to the imbalance - recovery mapping rule, and establish an adaptive mapping parameter based on the perturbation control parameters.

[0080] Specifically, transform the core imbalance control feature set into an operable control mapping rule, and establish a decision - making mechanism for imbalance control through pattern recognition and parameter association. This mapping rule can automatically select the optimal imbalance control strategy according to the current flight state, realizing intelligent adaptive control.

[0081] In some embodiments, establishing the imbalance-recovery mapping rule based on the core imbalance control feature set includes: identifying beneficial imbalance patterns on the core imbalance control feature set; extracting disturbance triggering characteristic parameters corresponding to the beneficial imbalance patterns; establishing an association relationship between the disturbance triggering characteristic parameters and the balance recovery index; and generating the imbalance-recovery mapping rule based on the association relationship.

[0082] Beneficial imbalance patterns were identified for the core imbalance control feature set. Principal component analysis was used to reduce the dimensionality of the feature set, extracting the main feature components with the highest contribution. The feature data was then grouped using the K-means clustering algorithm and hierarchical cluster analysis. The optimal number of clusters and cluster centers were determined by calculating the Euclidean distance and correlation coefficient between the features. Analysis of the synergistic relationship between speed control and precision control features revealed a significant positive synergistic effect when the speed control feature exhibited a high angular recovery rate and the precision control feature exhibited high error convergence accuracy. This led to the identification of a fast-response imbalance pattern, suitable for control scenarios requiring rapid and precise attitude adjustments within a short timeframe. Analysis of the matching relationship between oscillation suppression and coupling coordination features revealed that when the oscillation suppression feature exhibited an ideal damping ratio and the coupling coordination feature exhibited high multi-axis synchronicity, the control mechanism exhibited excellent stability. This led to the identification of a stable coordinated imbalance pattern, suitable for complex scenarios requiring control process smoothness and multi-axis coordination. By analyzing the fusion relationship between frequency and amplitude modulation features, we found that when the frequency modulation feature exhibits good frequency conversion smoothness and the amplitude modulation feature has high amplitude control accuracy, we can identify the adaptive modulation imbalance mode, which is specifically suitable for dynamic scenarios with changing environmental conditions and task mode switching. For example, in forest fire monitoring tasks, the rapid response mode is activated for rapid fire tracking, the stable coordination mode is activated for stable image acquisition, and the adaptive mode is activated when adjusting the observation strategy.

[0083] The disturbance trigger characteristic parameters corresponding to the beneficial imbalance mode are extracted. Based on the three identified beneficial imbalance modes, feature engineering and parameter extraction techniques are used to accurately extract key disturbance trigger parameters from the characteristic combinations of each mode. For the rapid response imbalance mode, statistical analysis and threshold optimization methods are used to extract the angle recovery rate threshold as a velocity trigger parameter, while the error convergence accuracy threshold is also extracted as an accuracy trigger parameter. A dual-parameter joint discrimination mechanism is established, and the mode is activated only when both parameters meet high-level requirements. For the stable coordinated imbalance mode, sensitivity analysis and interval optimization techniques are used to extract the damping ratio range as a stability trigger parameter, while the multi-axis synchronicity index is also extracted as a coordination trigger parameter. The mode is activated when both parameters are within the ideal value range. For the adaptive adjustment imbalance mode, frequency domain analysis and amplitude statistics are used to extract the frequency conversion rate as a frequency trigger parameter, and the amplitude adjustment range as an amplitude trigger parameter. The mode is activated based on these two parameters when environmental changes or mission switching requirements are detected. In special missions such as formation flying, the trigger parameters can be dynamically adjusted according to mission requirements.

[0084] A correlation between the disturbance-triggering characteristic parameters and the balance recovery index was established. Multivariate statistical analysis and regression modeling methods were used to establish a quantitative relationship between the parameters. A correlation analysis was conducted between the angle recovery rate threshold and the balance recovery efficiency index. Linear regression revealed a positive correlation between the two. A linear model was established: η_r = a·θ_th + b, where η_r represents the recovery efficiency index, θ_th represents the angle recovery rate threshold, and the regression coefficient a reflects the impact of threshold changes on efficiency. An optimization relationship model was established between the damping ratio range and the stability index: S_stab = -α(ζ - ζ_opt)² + S_max, where ζ represents the damping ratio, ζ_opt represents the optimal damping ratio, and S_max represents the maximum stability index. Curve fitting revealed that optimal stability is achieved when the damping ratio falls within a specific ideal range. A piecewise nonlinear relationship model was established between the frequency conversion rate and the adaptability index. Within a reasonable rate range, the control process maintains good adaptability, but performance significantly degrades when the critical value is exceeded. A strong correlation was established between the multi-axis synchronization index and the coordinated recovery index: higher synchronization indicates better coordinated recovery performance. There is an optimal balance point between the amplitude adjustment range and control accuracy, which needs to be optimized and determined.

[0085] Imbalance-recovery mapping rules are generated based on the aforementioned associations. Based on the established associations, rule-based reasoning and decision tree algorithms are used to generate executable mapping rules. Rapid response mapping rules: When the angle recovery rate requirement exceeds a threshold and the error convergence accuracy reaches a high standard, the target recovery efficiency and accuracy targets are automatically calculated based on the linear association model, the rapid response imbalance mode is activated, and the corresponding control parameters are set. Stable coordination mapping rules: When a multi-axis coordinated control requirement is detected and the damping ratio is within the ideal range, the optimal coordination parameters are calculated based on the optimization model, the stable coordination imbalance mode is activated, and the multi-axis coordination strategy is optimized. Adaptive mapping rules: When environmental conditions or mission modes change, the optimal transition strategy is determined based on the nonlinear relationship model, switching to the adaptive imbalance adjustment mode and dynamically adjusting the control parameters. The mapping rules integrate intelligent mode switching logic, enabling smooth and automatic switching between different modes based on flight conditions. In complex weather conditions, the mapping rules can automatically adjust the activation priority of each mode to enhance the control's anti-interference capability.

[0086] Disturbance control parameters are generated according to the imbalance-recovery mapping rule. Leveraging the linear correlation model and control target settings in the fast-response mapping rule, when the corresponding trigger condition is detected, the fast-response disturbance control parameters are calculated according to the mapping rule: disturbance amplitude δ = f1(η_target) = k_δ·η_target, disturbance duration T_d = f2(P_target) = T_base·(P_target / P_ref)^β, and recovery time constant τ_r = f3(a,b) = (a·θ_th +b)⁻¹, where η_target is the target recovery efficiency, P_target is the accuracy target, and k_δ, T_base, and β are configuration parameters. Leveraging the correlation relationships and coordination parameter calculations in the stable coordination mapping rule, when the multi-axis coordinated control conditions are met, the stable coordinated disturbance control parameters are generated according to the mapping rule. These parameters include the main axis disturbance configuration determined by the optimal coordination parameters, the auxiliary axis disturbance configuration based on the multi-axis coordination strategy, and the inter-axis timing parameters optimized based on the coordination relationship. Leveraging the nonlinear relationships and equilibrium point principles within the adaptive mapping rule, adaptive disturbance control parameters are generated according to the mapping rule when the environment or task changes. These parameters include frequency parameters determined by the optimal transition strategy, amplitude parameters set based on dynamic adjustment requirements, and transition time parameters optimized based on adaptability. For example, in a forest fire monitoring task, when the observation requirement switches from static monitoring to dynamic tracking, the control algorithm automatically adjusts the disturbance control parameters according to the rapid response mapping rule, switching from the 5° disturbance amplitude and 1.2s recovery time of the stable coordination mode to the 8° disturbance amplitude and 0.8s recovery time of the rapid response mode. Through parameterized transformation of the imbalance-recovery mapping rule, disturbance control parameters adapted to different control requirements are generated.

[0087] Adaptive mapping parameters are established based on the disturbance control parameters. Fast-response adaptive mapping parameters are established based on the disturbance amplitude, disturbance duration, and recovery time constant in the fast-response disturbance control parameters. These parameters serve as baseline values, setting an adaptive adjustment mechanism. When the actual control effect deviates from expectations, the disturbance amplitude and recovery time constant are automatically adjusted based on the degree of deviation, ensuring the stability of fast-response performance. Stable coordinated adaptive mapping parameters are established based on the primary and secondary axis disturbance configurations and inter-axis timing parameters in the stable coordinated disturbance control parameters. These parameters form the basis for a multi-axis adaptive mechanism. When changes in inter-axis coordination are detected, the disturbance configurations and timing parameters of the primary and secondary axes are automatically adjusted to maintain the coordinated control effect. Adaptive adjustment mapping parameters are established based on the frequency, amplitude, and transition time parameters in the adaptive disturbance control parameters. These parameters serve as the core of an environmental adaptation mechanism, dynamically adjusting the frequency and amplitude parameters based on the intensity of environmental disturbances and the degree of task changes, ensuring control adaptability in changing environments. Through adaptive expansion and intelligent optimization of disturbance control parameters, adaptive mapping parameters with environmental adaptability and learning capabilities are established.

[0088] Step S170 , generating an inverse imbalance joint control according to the adaptive mapping parameters, outputting a flight control signal according to the inverse imbalance joint control, and realizing adaptive imbalance control of the flight attitude of the UAV.

[0089] Specifically, a multi-channel imbalance control matrix is ​​first established based on adaptive mapping parameters. The three mapping parameters (rapid response, stable coordination, and adaptive adjustment) are assigned as weights to the three columns of the matrix, respectively. Three rows of the matrix are formed for the pitch, roll, and yaw axes, creating a 3×3 control matrix. Each element corresponds to the imbalance control weight of a specific axis in the corresponding mode. For example, when rapid adjustment of the pitch axis is required in forest fire monitoring, the first row and first column of the matrix are assigned a high weight. Next, an attitude coupling analysis is performed based on the control matrix to determine the inverse control gain for each axis. The dominant control mode for each axis is identified by extracting the maximum value from each row. The maximum weight is then used as the baseline for the inverse control gain of that axis and normalized. For example, if the pitch axis has the highest rapid response weight during payload release, its control gain is set accordingly. Then, the inverse control gain is adaptively fused with the adaptive mapping parameters to generate a joint control strategy. The enhanced control parameters for each axis are obtained by multiplying the gain by the corresponding mapping parameter. Control priorities are determined based on the gain, and coordination rules are established. For example, in formation flying, the highest priority is assigned to the pitch axis with the highest gain. Finally, based on the joint control strategy, the control output intensity is adjusted in real time to form a reverse imbalance joint control. The output intensity of each axis is dynamically adjusted according to the ratio of the actual deviation to the target deviation, and a total control mechanism is established to prevent overload. For example, when the yaw demand increases under complex weather conditions, the yaw output is moderately increased within the total limit and the intensity of other axes is reduced. A complete reverse imbalance joint control is formed through parameter application and proportional adjustment.

[0090] Based on the combined control of reverse imbalance, flight control signals are output to achieve adaptive imbalance control of the UAV's flight attitude. The actual output intensity of the pitch axis is directly converted into the elevator deflection angle. Pre-set conversion coefficients are used to map the output intensity to specific rudder surface deflection degrees and deflection rates, generating an elevator control signal. The actual output intensity of the roll axis is converted into aileron differential deflection angles. The differential coefficients are used to calculate the symmetrical deflection degrees of the left and right ailerons, generating an aileron differential control signal. The actual output intensity of the yaw axis is converted into a rudder deflection angle. The yaw conversion coefficients are used to determine the rudder deflection direction and amplitude, generating a rudder control signal. The sum of the three-axis output intensities is used to calculate the throttle opening increment using the power conversion coefficient, generating a throttle control signal that matches the imbalance control requirements. All control signals are amplitude-limited and safety-checked to ensure that the control instructions of each actuator are within the designed safety range. For example, in a target tracking mission, when the pitch axis output strength is 6.5, the roll axis output strength is 4.2, and the yaw axis output strength is 3.8, the conversion coefficients generate specific control instructions for the elevator to tilt upward by 13 degrees, the left aileron to tilt downward by 8.4 degrees, the right aileron to tilt upward by 8.4 degrees, the rudder to tilt right by 11.4 degrees, and the throttle opening to increase by 14.5%. The various actuators coordinate their actions according to these precise numerical instructions, achieving rapid and accurate attitude adjustment. By numerically converting the reverse imbalance joint control strength and precisely driving the actuators, adaptive imbalance control of the UAV's flight attitude is achieved.

[0091] In order to implement the UAV flight attitude adaptive imbalance control method corresponding to the above method embodiment, to achieve the corresponding functions and technical effects. Figure 2 , Figure 2 The following is a block diagram of a UAV flight attitude adaptive imbalance control device 200 provided in an embodiment of the present application. For ease of explanation, only the parts related to this embodiment are shown. The UAV flight attitude adaptive imbalance control device 200 provided in an embodiment of the present application includes:

[0092] The trajectory analysis module 201 is used to obtain the current attitude state and the target attitude state of the UAV, and perform reverse trajectory analysis on the current attitude state and the target attitude state to generate an ideal control trajectory sequence;

[0093] An imbalance acquisition module 202 is configured to acquire an imbalance point of the ideal control trajectory sequence, determine a disturbance injection position according to the imbalance point, and generate a controllable imbalance strategy library according to the disturbance injection position;

[0094] A benchmark establishment module 203 is configured to determine an imbalance control interval based on the ideal control trajectory sequence and the controllable imbalance strategy library, generate a disturbance injection sequence based on the imbalance control interval, and establish a dynamic balance benchmark based on the disturbance injection sequence;

[0095] A trajectory reconstruction module 204 is configured to perform trajectory reconstruction based on the dynamic balance reference to obtain an imbalance-recovery control trajectory, and obtain a dynamic response feature set corresponding to the imbalance-recovery control trajectory;

[0096] A feature extraction module 205 is configured to obtain a balance recovery index corresponding to the dynamic response feature set, construct an imbalance recovery spatiotemporal manifold based on the balance recovery index, and perform feature mapping on the imbalance recovery spatiotemporal manifold to obtain a core imbalance control feature set;

[0097] a mapping generation module 206 for establishing an imbalance-recovery mapping rule according to the core imbalance control feature set, generating disturbance control parameters according to the imbalance-recovery mapping rule, and establishing adaptive mapping parameters based on the disturbance control parameters;

[0098] The control execution module 207 is used to generate an inverse imbalance joint control according to the adaptive mapping parameters, output a flight control signal according to the inverse imbalance joint control, and realize adaptive imbalance control of the UAV flight attitude.

[0099] The aforementioned unmanned aerial vehicle (UAV) flight attitude adaptive imbalance control device 200 can implement the UAV flight attitude adaptive imbalance control method of the aforementioned method embodiment. The optional options in the aforementioned method embodiment also apply to this embodiment and will not be described in detail here. The remaining contents of the present application embodiment can be referenced to the contents of the aforementioned method embodiment and will not be further described in this embodiment.

[0100] like Figure 3 As shown, the third embodiment of the present invention further provides a computer device, including a memory 301, a processor 302, and a computer program stored in the memory 301 and executable on the processor 302, characterized in that when the processor 302 executes the program, the steps of the method for adaptive imbalance control of the flight attitude of a drone described in the first embodiment of the present invention are implemented.

[0101] The purpose of the above embodiments is to exemplify and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosed content of the present invention, and it does not limit the scope of protection of the present invention.

[0102] The above embodiments are not exhaustive and may include many other embodiments not listed above. Any replacements and improvements made without violating the concept of the present invention are within the scope of protection of the present invention.

Claims

1. A method for adaptive unbalance control of UAV flight attitude, characterized in that: include: Acquiring the current attitude state and the target attitude state of the UAV, and performing inverse trajectory analysis on the current attitude state and the target attitude state to generate an ideal control trajectory sequence; Obtaining an imbalance point of the ideal control trajectory sequence, determining a disturbance injection position according to the imbalance point, and generating a controllable imbalance strategy library according to the disturbance injection position; determining an imbalance control interval according to the ideal control trajectory sequence and the controllable imbalance strategy library, generating a disturbance injection sequence according to the imbalance control interval, and establishing a dynamic balance benchmark according to the disturbance injection sequence; Reconstructing a trajectory based on the dynamic balance reference to obtain an imbalance-recovery control trajectory, and obtaining a dynamic response feature set corresponding to the imbalance-recovery control trajectory; Obtaining a balance recovery index corresponding to the dynamic response feature set, constructing an imbalance recovery space-time manifold based on the balance recovery index, and performing feature mapping on the imbalance recovery space-time manifold to obtain a core imbalance control feature set; establishing an imbalance-recovery mapping rule according to the core imbalance control feature set, generating a disturbance control parameter according to the imbalance-recovery mapping rule, and establishing an adaptive mapping parameter based on the disturbance control parameter; An inverse imbalance joint control is generated according to the adaptive mapping parameters, and a flight control signal is output according to the inverse imbalance joint control to achieve adaptive imbalance control of the flight attitude of the UAV.

2. The method according to claim 1, characterized in that The performing reverse trajectory analysis on the current posture state and the target posture state to generate an ideal control trajectory sequence includes: Perform reverse path planning from the target posture state to the current posture state; Determine an intermediate control node according to the reverse path planning, and perform time sequence marking on the intermediate control node; Obtaining a posture change gradient of the intermediate control node; An ideal control trajectory sequence is generated according to the timing mark and the posture change gradient.

3. The method according to claim 1, characterized in that The obtaining of the imbalance point of the ideal control trajectory sequence includes: Analyzing the key points of posture conversion in the ideal control trajectory sequence; determining an available natural imbalance direction at the posture transition key point; evaluating the controllability of the natural imbalance direction and generating a controllability evaluation result; An imbalance point is obtained according to the natural imbalance direction and the controllability evaluation result.

4. The method according to claim 1, wherein Generating a controllable imbalance strategy library according to the disturbance injection position includes: Establishing a corresponding relationship between disturbance amplitude and recovery time for each disturbance injection position; Determining an optimal disturbance parameter combination according to the corresponding relationship; The optimal disturbance parameter combinations are classified and stored to generate a controllable imbalance strategy library.

5. The method according to claim 1, characterized in that The establishing of a dynamic balance benchmark according to the disturbance injection sequence comprises: determining an expected recovery trajectory after the imbalance based on the disturbance injection sequence; generating a tolerance range based on the expected recovery trajectory; determining a time reference point according to the expected recovery trajectory; The expected recovery trajectory, the tolerance range, and the time reference point are combined to form a dynamic balance reference.

6. The method according to claim 1, characterized in that The step of reconstructing a trajectory based on the dynamic balance reference to obtain an imbalance-recovery control trajectory includes: Performing back-calculation of the trajectory during the imbalance phase according to the dynamic balance reference to generate a back-calculated trajectory segment; determining an imbalance-recovery transition node based on the inverse trajectory segment; Trajectory splicing is performed according to the conversion node and the dynamic balance reference to generate an imbalance-recovery control trajectory.

7. The method according to claim 1, characterized in that The establishing of an imbalance-recovery mapping rule according to the core imbalance control feature set includes: performing beneficial imbalance pattern identification on the core imbalance control feature set; extracting disturbance triggering characteristic parameters corresponding to the beneficial imbalance mode; Establishing a correlation relationship between the disturbance trigger characteristic parameter and the balance recovery index; An imbalance-recovery mapping rule is generated according to the association relationship.

8. The method according to claim 1, characterized in that Generating the reverse imbalance joint control according to the adaptive mapping parameters includes: establishing a multi-channel imbalance control matrix based on the adaptive mapping parameters; Performing attitude coupling analysis based on the multi-channel imbalance control matrix to determine the inverse control gain of each axis; Adaptively fusing the inverse control gain with the adaptive mapping parameter to generate a joint control strategy; Based on the joint control strategy, the control output intensity is adjusted in real time to form reverse imbalance joint control.

9. A UAV flight attitude adaptive imbalance control device, characterized in that: include: A trajectory analysis module is used to obtain the current attitude state and the target attitude state of the UAV, and perform reverse trajectory analysis on the current attitude state and the target attitude state to generate an ideal control trajectory sequence; an imbalance acquisition module, configured to acquire an imbalance point of the ideal control trajectory sequence, determine a disturbance injection position according to the imbalance point, and generate a controllable imbalance strategy library according to the disturbance injection position; a benchmark establishment module, configured to determine an imbalance control interval according to the ideal control trajectory sequence and the controllable imbalance strategy library, generate a disturbance injection sequence according to the imbalance control interval, and establish a dynamic balance benchmark according to the disturbance injection sequence; a trajectory reconstruction module, configured to perform trajectory reconstruction based on the dynamic balance reference to obtain an imbalance-recovery control trajectory, and obtain a dynamic response feature set corresponding to the imbalance-recovery control trajectory; a feature extraction module, configured to obtain a balance recovery index corresponding to the dynamic response feature set, construct an imbalance recovery spatiotemporal manifold based on the balance recovery index, and perform feature mapping on the imbalance recovery spatiotemporal manifold to obtain a core imbalance control feature set; a mapping generation module, configured to establish an imbalance-recovery mapping rule according to the core imbalance control feature set, generate a disturbance control parameter according to the imbalance-recovery mapping rule, and establish an adaptive mapping parameter based on the disturbance control parameter; The control execution module is used to generate an inverse imbalance joint control according to the adaptive mapping parameters, output a flight control signal according to the inverse imbalance joint control, and realize adaptive imbalance control of the flight attitude of the UAV.

10. A computer device, characterized in that: The method comprises a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the method according to any one of claims 1 to 8 when executing the computer program.

Images