Fixed-wing unmanned aerial vehicle six-degree-of-freedom sliding mode control method based on all-drive modeling and medium

Abstract

The invention belongs to the technical field of aircraft flight control and intelligent robust control, and particularly relates to a fixed-wing unmanned aerial vehicle six-degree-of-freedom sliding mode control method based on all-drive modeling and a medium. Constructing an all-wheel-drive control input vector according to a translational channel control quantity (including equivalent thrust, obtained through decomposition of comprehensive thrust of the unmanned aerial vehicle, of an inertial system in each axial direction) and a rotating channel control quantity (including control surface deflection angles of ailerons, elevators and rudders of the unmanned aerial vehicle); constructing a six-degree-of-freedom state vector according to the position attitude state (including the translation amount and the rotation amount of the unmanned aerial vehicle in each axial direction of the inertial system) of the unmanned aerial vehicle under the inertial system, and constructing an affine all-drive model according to the all-drive control input vector, the six-degree-of-freedom state vector and the input gain matrix; six-channel unified modeling and decoupling regulation are realized, so that trajectory tracking precision, robustness and energy consumption suppression level are improved.

Description

Technical Field

[0001] This invention belongs to the field of aircraft flight control and intelligent robust control technology, specifically involving a six-degree-of-freedom sliding mode control method and medium for fixed-wing UAVs based on all-drive modeling. Background Technology

[0002] In recent years, fixed-wing unmanned aerial vehicles (UAVs) have been widely used in low-altitude economic operations, emergency rescue, and large-scale inspections due to their advantages such as high cruising speed, long range, and high energy efficiency. However, the dynamics of fixed-wing aircraft naturally exhibit an underactuated structure, with the actual number of control inputs being less than the system's degrees of freedom. This results in strong nonlinear coupling and channel constraints between translation and attitude. When performing large maneuvers, fault conditions, saturation constraints, or complex disturbances, a chain reaction of enhanced coupling, decreased controllability margin, and squeezed robustness margin can easily occur, significantly limiting closed-loop accuracy and stability margin.

[0003] From a modeling perspective, existing studies typically explicitly represent aerodynamic effects through lift, drag, lateral forces and their moment coefficients, angle of attack, sideslip angle dynamics, and aerodynamic derivatives. However, "considering aerodynamics" is not equivalent to "full-drive modeling." Under traditional underdriven input structures, even if the model contains relatively complete aerodynamic terms, thrust and control surfaces still act on the system only through limited physical channels. The input and output drive matrices of the model may still be rank deficient or irreversible in structure, making it difficult to achieve unified decoupled control and isomorphic design across all degrees of freedom.

[0004] Furthermore, the aerodynamic terms, with dynamic pressure as the main scale, permeate both the aerodynamic force and torque channels, making the control of fixed wings more structurally challenging within the full envelope: First, aerodynamic forces and torques generally satisfy a scaling relationship related to the square of velocity, resulting in strong time-varying time-dependent changes in the equivalent input gain from control surface deflection angle and thrust to acceleration. The same control surface deflection angle is amplified by dynamic pressure at high speeds and significantly attenuated at low speeds, and is further modulated by angle of attack and sideslip angle. Second, the aerodynamic effects of the control surfaces appear as force rectangles in the rotational equations and can also be projected into the translational channel through the translational equations, manifesting as enhanced cross-channel coupling and deterioration of the input mapping condition number. Third, actuator saturation and rate and bandwidth limitations make it difficult to simply cover the strong time-varying gain by increasing the control force. At high speeds, structural amplification of switching terms can easily lead to oscillations and energy waste, while at low speeds, saturation and hysteresis are more likely to be triggered, resulting in slow response or even decoupling degradation.

[0005] Therefore, existing technologies require a six-degree-of-freedom sliding mode control method for UAVs that is isomorphically compatible with the all-drive model. Summary of the Invention

[0006] To address the problems existing in the prior art, the present invention aims to provide a six-degree-of-freedom sliding mode control method and medium for fixed-wing UAVs based on all-drive modeling. The method constructs an all-drive control input vector based on translational control quantities (including equivalent thrust along each axis of the inertial frame obtained from the comprehensive thrust decomposition of the UAV) and rotational control quantities (including the surface deflection angles of the UAV's ailerons, elevators, and rudders). A six-degree-of-freedom state vector is constructed based on the UAV's position and attitude state in the inertial frame (including translational and rotational quantities along each axis of the inertial frame). An affine all-drive model is constructed based on the all-drive control input vector, the six-degree-of-freedom state vector, and the input gain matrix, achieving unified modeling and decoupling control across the six channels, thereby improving trajectory tracking accuracy, robustness, and energy consumption suppression.

[0007] In a first aspect, the present invention provides a six-degree-of-freedom sliding mode control method for a fixed-wing unmanned aerial vehicle based on all-drive modeling, comprising: Based on the aerodynamic model of the UAV in the airframe and the coordinate transfer matrix from the airframe to the inertial frame, the gain matrix of the effect of the rotation channel control quantity on the translational acceleration of the UAV in each axis of the inertial frame is determined; the rotation channel control quantity includes the surface deflection angles of the UAV's ailerons, elevators and rudders. The all-drive control input vector is constructed based on the translational and rotational control variables; the translational control variables include the equivalent thrust in each axis of the inertial frame obtained by the comprehensive thrust decomposition of the UAV. A six-degree-of-freedom state vector is constructed based on the position and attitude state of the UAV in the inertial frame; an input gain matrix is ​​constructed between the all-drive control input vector and the six-degree-of-freedom state vector based on the aforementioned gain matrix; An affine all-drive model is constructed based on the all-drive control input vector, the six-degree-of-freedom state vector, and the input gain matrix. Based on the desired position and attitude state, construct the position and attitude tracking error, and construct the sliding surface based on the position and attitude tracking error; An all-drive control structure based on an affine all-drive model and a sliding surface is constructed to generate the all-drive control input vector. Based on the all-drive control structure, the translational and rotational control variables are solved to achieve UAV control.

[0008] Optionally, the expression for the affine all-drive model is: ; ; in, Indicates the position and attitude state; Represents a six-degree-of-freedom state vector; The second derivative of the position and attitude state; Indicates the drift term; Represents the input gain matrix; This represents the all-wheel drive control input vector; Indicates the combined disturbance term; , and These represent the translational moments of the UAV along the x, y, and z axes in the inertial frame, respectively. , , These represent the translational accelerations of the UAV along the x, y, and z axes in the inertial frame, respectively. , and These represent the rotations of the UAV around the x-axis, y-axis, and z-axis in the inertial frame, respectively. , , They represent , and The second derivative of ; and They represent and The first derivative of; , , These represent the deterministic components related to the translational momentum of the UAV along the x, y, and z axes of the inertial frame, respectively. and The first x-axis rudder coupling gain and the second x-axis rudder coupling gain represent the amount of rotation of the UAV around the x-axis in the inertial frame, respectively. , and The first y-axis rudder coupling gain, the second y-axis rudder coupling gain, and the third y-axis rudder coupling gain represent the amount of rotation of the UAV around the y-axis in the inertial frame, respectively. and The first z-axis rudder coupling gain and the second z-axis rudder coupling gain represent the amount of rotation of the UAV around the z-axis in the inertial frame, respectively. Indicates the quality of the drone; , and These represent the linear contribution vectors of the ailerons, elevator, and rudder of the UAV, respectively. Indicates air dynamic pressure. Indicates the reference area; Represents the coordinate transition matrix from the machine system to the inertial system. The Row vectors; , and These represent the equivalent thrust along the x-axis, y-axis, and z-axis of the inertial frame obtained from the comprehensive thrust decomposition of the UAV, respectively. , and These represent the deflection angles of the ailerons, elevator, and rudder of the UAV, respectively. , , These represent the perturbation terms of the translational motion of the UAV along the x, y, and z axes of the inertial frame, respectively. , , These represent disturbance terms indicating the amount of rotation of the UAV along the x, y, and z axes of the inertial frame, respectively.

[0009] Optionally, , , The translational drift term constituted The expression is: ; ; ; in, Represents the translational state vector The corresponding translational drift term; Indicates air density; Indicates airspeed; Indicates the quality of the drone; , and These represent the x-axis, y-axis, and z-axis reference aerodynamic coefficients obtained by mapping the uncontrollable aerodynamic components of the lift, drag, and lateral force aerodynamic coefficients onto the machine system, respectively. , and These represent the uncontrollable aerodynamic components of the aerodynamic coefficients for lift, drag, and lateral force, respectively. Indicates the angle of attack; , and These represent the roll angular velocities of the UAV in the machine system. Pitch angular velocity With yaw rate dimensional value; , and These represent the basic coefficients for lift, drag, and lateral force, respectively. and They represent the angle of attack. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Indicates the sideslip angle The corresponding lateral force aerodynamic coefficient; and They represent pitch angular velocities, respectively. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Represents roll angular velocity The corresponding lateral force aerodynamic coefficient; Indicates yaw rate The corresponding lateral force aerodynamic coefficient.

[0010] Optionally, the expression for the linear contribution vectors of the UAV's ailerons, elevator, and rudder is: ; in, Indicates the angle of attack; Indicates the aileron deflection angle The corresponding lateral force aerodynamic coefficient; and These represent the elevator surface deflection angles. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Indicates the rudder surface deflection angle The corresponding lateral force aerodynamic coefficient.

[0011] Optionally, the expression for the sliding surface is: ; ; in, Indicates the sliding surface. Indicates the position and attitude tracking error. The derivative of the position and attitude tracking error; Indicates the first An exponential adaptive gain Indicates the first The derivative of an exponential adaptive gain; Represents a symbolic function. and These represent the first and second nonlinear correction orders, respectively. The power exponent of the adaptive law; Indicates the first The lower bound of the gain for an exponential adaptive gain; Indicates the first A fallback factor for an exponential adaptive gain. Indicates the first The growth rate coefficient of the exponential adaptive gain.

[0012] Optionally, the expression for the all-drive control structure is: ; in, Indicates position and attitude state The second derivative of the expected state; The matrix representing the inverse of the input gain matrix; Indicates the boundary layer thickness; express Adaptive robust gain at time step; This represents the hyperbolic tangent function.

[0013] Optionally, adaptive robust gain The derivative The expression is: ; in, Indicates the sliding mode threshold; Indicates the reference gain; , and These represent the first coefficient, the second coefficient, and the third coefficient, respectively. , and All are greater than 0.

[0014] Optionally, the translational and rotational control variables are solved using a modular decoupled solution method, and the solution steps include: All-drive control input vector Decomposed into a translational control input vector consisting of translational channel control quantities. and the rotation control input vector composed of rotation channel control quantities ; Input gain matrix Denoted in block form as , , and Let these represent the first sub-gain matrix, the second sub-gain matrix, and the third sub-gain matrix, respectively. The second sub-gain matrix... The gain matrix represents the effect of the rotation channel control quantity on the translational acceleration of the UAV in each axis of the inertial frame. The affine all-wheel drive model is decomposed into a translational model and a rotational model. The expressions for the translational and rotational models are as follows: ; in, The translational state vector, Let be the rotational state vector. and Let represent the second derivatives of the translational state vector and the rotational state vector, respectively; and These are the drift terms corresponding to the translational and rotational state vectors, respectively. and These are the combined disturbance terms corresponding to the translational state vector and the rotational state vector, respectively; Translational control input vector and rotation control input vector The solution result is expressed as: ; ; in, Represents the translational state vector The second derivative of the expected state, Represents the rotational state vector The second derivative of the expected state; This represents the rotation control input vector at the previous moment. The inverse matrix of the first sub-gain matrix; The inverse matrix of the third sub-gain matrix; This represents the sliding surface of the translational subsystem. Indicates the sliding surface of the rotating subsystem; The expression for the sliding surface of the translational subsystem is: in, This indicates the translational position tracking error. The derivative of the translational position tracking error is expressed. and These represent the first and second translational nonlinear correction orders, respectively. Represents a symbolic function; This represents the diagonal gain matrix of the low-power error terms in the translational three-channel system. , and These represent the first x-axis translational gain, the first y-axis translational gain, and the first z-axis translational gain, respectively. The diagonal gain matrix represents the high-power error terms of the translational three-channel circuit. , and These represent the second x-axis translational gain, the second y-axis translational gain, and the second z-axis translational gain, respectively. The expression for the sliding surface of the rotating subsystem is: in, Indicates the rotational attitude tracking error. The derivative of the rotational attitude tracking error; and These represent the first and second orders of rotational nonlinear correction, respectively. Let be the diagonal gain matrix of the three-channel low-power error terms, where , and These represent the first x-axis angle gain, the first y-axis angle gain, and the first z-axis angle gain, respectively. To rotate the diagonal gain matrix of the high-power error terms of the three channels, , and These represent the second x-axis angle gain, the second y-axis angle gain, and the second z-axis angle gain, respectively. Represents a symbolic function.

[0015] Optionally, the combined perturbation term of the affine all-drive model is observed using an enhanced perturbation observer, the expression of which is: ; ; ; ; ; ; in, Indicates time, This represents the output of the enhanced disturbance observer used to approximate the integrated disturbance term. Disturbance compensation item, and These represent the first adaptive gain adjustment and the second adaptive gain adjustment of the enhanced disturbance observer, respectively. Represents the observed sliding mode variable; This indicates the base gain for enhancing the perturbation observer; and These represent the normalized rate of change of the disturbance and the normalized energy density of the disturbance, respectively. Indicates a superspiral assisted state; The derivative of the superspinning auxiliary state; Indicates a positive coefficient; This indicates the base gain of the enhanced perturbation observer. This represents the comprehensive index of the response of the disturbed structure. Represents the structural response gain coefficient. This represents the coefficient of coupling relationship at the same scale. Indicates the switching function; This indicates the corresponding position and attitude state. The generalized pose state vector, The derivative of the generalized pose state vector is represented by . Represents the generalized velocity state vector. This represents the integrated control input vector. The derivative of the generalized velocity state vector is given by: As an auxiliary variable, The derivative of the auxiliary variable; After introducing the disturbance compensation term from the enhanced disturbance observer output into the all-drive control structure, the expression of the all-drive control structure is adjusted to: .

[0016] In a second aspect, the present invention provides a computer-readable storage medium storing a computer program for executing the six-degree-of-freedom sliding mode control method for a fixed-wing UAV based on all-drive modeling.

[0017] Based on the above technical solutions, the beneficial effects of this invention are as follows: This invention addresses the underactuation problem caused by the strong nonlinear coupling between attitude angle dynamics and translational motion dynamics due to insufficient control input dimensions in traditional fixed-wing UAVs. This invention proposes a high-order full-dimensional modeling method based on the theory of all-drive systems and the influence of aerodynamic control. This invention determines the gain matrix of the effect of rotational channel control quantities on the translational acceleration of the UAV in each axis of the inertial frame based on the aerodynamic model of the UAV in the inertial frame and the coordinate transfer matrix from the inertial frame to the machine frame. The rotational channel control quantities include the surface deflection angles of the UAV's ailerons, elevators, and rudders. A total drive control input vector is constructed based on the translational and rotational channel control quantities. The translational channel control quantities include the equivalent thrust in each axis of the inertial frame obtained from the comprehensive thrust decomposition of the UAV. A six-degree-of-freedom state vector is constructed based on the UAV's position and attitude state in the inertial frame. An input gain matrix between the total drive control input vector and the six-degree-of-freedom state vector is constructed based on the gain matrix. An affine total drive model is constructed based on the total drive control input vector, the six-degree-of-freedom state vector, and the input gain matrix. This model can achieve decoupling and coupling processing of the translational and rotational quantities of the UAV in each axis of the inertial frame, and maintains low-coupling control characteristics based on the invertible input gain matrix, effectively solving the problem of insufficient control dimensions in underactuated systems. This invention also constructs a position and attitude tracking error based on the desired position and attitude state, and constructs a sliding surface based on the position and attitude tracking error; constructs an all-drive control structure based on the affine all-drive model and the sliding surface, and solves for the translational channel control quantity and the rotational channel control quantity based on the all-drive control structure to achieve UAV control. Attached Figure Description

[0018] Figure 1 This is a flowchart of a six-degree-of-freedom sliding mode control method for a fixed-wing UAV based on all-drive modeling, according to an embodiment of the present invention. Figure 2 This is a comparison diagram of three-dimensional spatial trajectory tracking in the cyclic trajectory mode of this invention. Figure 3 This is a six-channel state tracking diagram in the cyclic trajectory mode of this invention embodiment; Figure 3 (a) is Channel position tracking diagram; Figure 3 (b) is Channel position tracking diagram; Figure 3 (c) is Channel position tracking diagram; Figure 3 (d) is Channel attitude angle tracking diagram; Figure 3 (e) is Channel attitude angle tracking diagram; Figure 3 (f) is Channel attitude angle tracking diagram; Figure 4 This is a comparison chart of the six-channel absolute errors in the cyclic trajectory mode of this invention. Figure 4 (a) is Comparison chart of absolute errors in channel position; Figure 4 (b) is Comparison chart of absolute errors in channel position; Figure 4 (c) is Comparison chart of absolute errors in channel position; Figure 4 (d) is Comparison chart of absolute error of channel attitude angle; Figure 4 (e) is Comparison chart of absolute error of channel attitude angle; Figure 4 (f) is Comparison chart of absolute error of channel attitude angle; Figure 5 This is a comparison chart of disturbance observation results in an embodiment of the present invention; Figure 5 (a) is Comparison chart of channel disturbance observations; Figure 5 (b) is Comparison chart of channel disturbance observations; Figure 5 (c) is Comparison chart of channel disturbance observations; Figure 5 (d) is Comparison chart of channel disturbance observations; Figure 5 (e) is Comparison chart of channel disturbance observations; Figure 5 (f) is Comparison chart of channel disturbance observations; Figure 6 This is a comparison diagram of control inputs in the cyclic trajectory mode according to an embodiment of the present invention; Figure 6 (a) represents the equivalent thrust. Control input comparison chart; Figure 6(b) represents the equivalent thrust. Control input comparison chart; Figure 6 (c) represents the equivalent thrust. Control input comparison chart; Figure 6 (d) represents the combined thrust. Control input comparison chart; Figure 7 This is a comparison diagram of actuator control surface commands in the cyclic trajectory mode of this invention. Figure 7 (a) represents the aileron deflection angle. Comparison diagram of control surface commands; Figure 7 (b) represents the elevator surface deflection angle. Comparison diagram of control surface commands; Figure 7 (c) represents the rudder surface deflection angle. A comparison diagram of control surface commands. Detailed Implementation

[0019] The technical solution of the present invention will be further described below. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of protection.

[0020] Combination Figure 1 This embodiment provides a six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling, which includes: Step S1: Construct an affine all-drive model.

[0021] Step S1 specifically includes: determining the gain matrix of the effect of the rotation channel control quantity on the translational acceleration of the UAV in each axis of the inertial frame based on the aerodynamic model of the UAV in the airframe and the coordinate transfer matrix from the airframe to the inertial frame; the rotation channel control quantity includes the surface deflection angles of the UAV's ailerons, elevators, and rudders; constructing the all-drive control input vector based on the translational channel control quantity and the rotation channel control quantity; the translational channel control quantity includes the equivalent thrust in each axis of the inertial frame obtained by the comprehensive thrust decomposition of the UAV; constructing a six-degree-of-freedom state vector based on the position and attitude state of the UAV in the inertial frame; constructing the input gain matrix between the all-drive control input vector and the six-degree-of-freedom state vector based on the gain matrix; and constructing an affine all-drive model based on the all-drive control input vector, the six-degree-of-freedom state vector, and the input gain matrix.

[0022] This embodiment uses a fixed-wing UAV at an angle of attack. Sideslip angle Based on the aerodynamic characteristics under control surface deflection conditions (aileron, elevator, and rudder), the dimensionless aerodynamic coefficients of lift, drag, and lateral force are expressed as a function of flight state (angle of attack). Sideslip angle Angular velocity (roll angular velocity in the machine system) Pitch angular velocity and yaw rate ) and control surface deflection (aileron control surface deflection) Elevator surface deflection rudder deflection angle A linear combination of ) can reduce modeling complexity while ensuring physical interpretability.

[0023] To account for the impact of angular velocity on aerodynamics, this embodiment specifies the roll angular velocity of the UAV. Pitch angular velocity With yaw rate After dimensionless processing, the expression is: ; in, , and These are the roll angular velocities. Pitch angular velocity With yaw rate The dimensionless value; in the above dimensionless processing, For wingspan, For the average aerodynamic chord length, For airspeed, dimensionless processing is used to incorporate the influence of angular velocity into the aerodynamic model in a scale-consistent manner.

[0024] Furthermore, this embodiment decomposes the aerodynamic coefficients into "controllable terms" directly modulated by the control surface deflection angle and "uncontrollable terms" determined by flight state and angular velocity. The aerodynamic coefficients of lift, drag, and lateral force are mapped onto the engine system via geometric projection to obtain equivalent aerodynamic coefficients. This allows the effect of the control surface deflection angle on the translational channel to be extracted in explicit gain form, laying the foundation for the subsequent construction of an all-wheel drive control input vector jointly driven by equivalent thrust and control surface deflection angle. The aerodynamic model is expressed as follows: ; In the above expression, , , These represent the aerodynamic coefficients for lift, drag, and lateral force, respectively. , and These represent the deflection angles of the aileron, elevator, and rudder, respectively. , and These are the basic coefficients representing the aerodynamic coefficients for lift, drag, and lateral force, respectively. and They represent the angle of attack. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Indicates the sideslip angle The corresponding lateral force aerodynamic coefficient; and They represent pitch angular velocities, respectively. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Represents roll angular velocity The corresponding lateral force aerodynamic coefficient; Indicates yaw rate The corresponding lateral force aerodynamic coefficient; Indicates the aileron deflection angle The corresponding lateral force aerodynamic coefficient; and These represent the elevator surface deflection angles. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; This represents the aerodynamic coefficient of the lateral force corresponding to the deflection angle of the rudder surface.

[0025] Angle of attack With sideslip angle The expression is: ; in, Represents the arctangent function; Represents the arcsine function; , and These represent the velocity components of the UAV on the x, y, and z axes in the machine system.

[0026] Aerodynamic coefficients for lift, drag, and lateral forces , and To separate them, the expression is: ; The aforementioned separation ensures that the "controllable term" is generated solely by the rudder surface deflection angle, while , and These represent the uncontrollable aerodynamic components of the aerodynamic coefficients for lift, drag, and lateral force, respectively. The uncontrollable aerodynamic components are determined by flight conditions, angular velocity, and other relevant influencing factors.

[0027] , and The expression is: ; in, , and Characterizing the baseline aerodynamic contribution under a given state facilitates the subsequent extraction of the contribution of the control surface deflection angle to the translation of the UAV in the form of a linear gain.

[0028] Will , and The expression mapping to the machine system is: ; in, Indicates the angle of attack; , and The uncontrollable aerodynamic components representing the lift, drag, and lateral force coefficients are mapped to the x-axis, y-axis, and z-axis reference aerodynamic coefficients of the engine system (excluding control surface deflection increments). This mapping allows the aerodynamic coefficients of lift, drag, and lateral force to be correlated with the attitude angle (angle of attack). Coupled and maintain a resolvable structure.

[0029] Furthermore, , and The expressions for the total equivalent aerodynamic coefficients along the x-axis, y-axis, and z-axis of the mechanical system after accounting for the control surface deflection increment are as follows: ; Multiply the above total equivalent aerodynamic coefficient by the dynamic pressure. and reference area The magnitude of the aerodynamic force on the machine system can be obtained. Then, by projecting the machine system onto the inertial frame, an aerodynamic force term can be added to the translational motion equations, expressed as: ; in, , and These represent the translational moments of the UAV along the x, y, and z axes in the inertial frame, respectively. , , These represent the translational accelerations of the UAV along the x, y, and z axes in the inertial frame, respectively. and These represent the rotations of the UAV around the y-axis and z-axis in the inertial frame, respectively. Indicates the quality of the drone. Represents gravitational acceleration. Indicates combined thrust. This represents the external disturbance vector normalized to acceleration; Represents the coordinate transition matrix The Row vectors.

[0030] coordinate transition matrix The expression is: ; in, , , Represent the rotation of the UAV around the x-axis, y-axis, and z-axis in the inertial frame, respectively; coordinate transition matrix. Used to describe the projection relationship from a mechanical system to an inertial frame.

[0031] This embodiment constructs a unified dynamics framework for UAVs in an inertial frame, incorporating both translational and rotational channels that include aerodynamic influences. This framework integrates the traditional approach that relies solely on the overall thrust of the UAV. The translational input of the drive extends into equivalent thrust along the x, y, and z axes of the inertial frame. , and Furthermore, the linear contribution of the control surface deflection angle to the aerodynamic coefficient is explicitly injected into the translational equations of the UAV in the inertial frame through attitude projection, thereby enabling the UAV to obtain directly adjustable control input channels in both the three translational and three rotational degrees of freedom in the inertial frame. In this modeling process, this embodiment uses dynamic pressure... Characterizing aerodynamic amplitude, air density, Airspeed, with reference area With coordinate transformation matrix The aerodynamic forces of the machine system are projected onto the inertial frame of reference, and external unmodeled disturbances and wind field interference are integrated into the comprehensive disturbance term. Ultimately, we obtain the following: The affine all-wheel drive model of the structure, in which, Indicates the position and attitude state; Indicates the drift term; Represents the input gain matrix; This represents the all-wheel drive control input vector; This represents the combined disturbance term, thus forming a "6 inputs and 6 outputs" total drive control input.

[0032] Combined thrust The translational input of the drive is extended into equivalent thrust in the three axes of the inertial frame. , and The expression is: ; in, For the quality of drones, For gravitational acceleration, this relationship is used to ensure that the amplitude of the equivalent thrust in the three axes is consistent with the combined thrust of the engine.

[0033] , and The expression is: ; in, and These represent the rotation of the UAV around the y-axis and z-axis in the inertial frame, respectively. This decomposition couples the thrust direction with the attitude angle, enabling the three-axis acceleration of the inertial frame to obtain an equivalent and independent thrust input form.

[0034] The expression for the linear contribution vectors of the ailerons, elevator, and rudder of the UAV is: ; in, , , These are the linear contribution vectors of the ailerons, elevator, and rudder of the UAV, respectively, used to explicitly express the effect of the control surface deflection angle on the translational channel in the form of "input gain" in order to facilitate the construction of an affine all-drive model.

[0035] The expression for the affine all-drive model constructed in this embodiment is: ; ; in, Indicates the position and attitude state; This represents a six-degree-of-freedom state vector. The second derivative of the position and attitude state; Represents the input gain matrix; This represents the all-wheel drive control input vector; Indicates the combined disturbance term; Indicates the quality of the drone; Indicates air dynamic pressure. Indicates the reference area; , , These represent the second derivatives of the translational moments of the UAV along the x, y, and z axes in the inertial frame, respectively. , and These represent the rotations of the UAV around the x-axis, y-axis, and z-axis in the inertial frame, respectively. , , These represent the second derivatives of the rotations of the UAV around the x-axis, y-axis, and z-axis in the inertial frame, respectively. and They represent and The first derivative of; , and These represent the equivalent thrust along the x-axis, y-axis, and z-axis in the inertial frame, obtained from the comprehensive thrust decomposition of the UAV, respectively. , and These represent the deflection angles of the ailerons, elevator, and rudder of the UAV, respectively. This represents the combined disturbance term caused by external gusts, model uncertainties and unmodeled coupling, and other influencing factors; , and These represent the perturbation terms of the translational motion of the UAV along the x, y, and z axes, respectively. , and These represent disturbance terms indicating the amount of rotation of the UAV along the x-axis, y-axis, and z-axis, respectively. and The first x-axis rudder coupling gain and the second x-axis rudder coupling gain represent the amount of rotation of the UAV around the x-axis in the inertial frame, respectively. , and The first y-axis rudder coupling gain, the second y-axis rudder coupling gain, and the third y-axis rudder coupling gain represent the amount of rotation of the UAV around the y-axis in the inertial frame, respectively. and The first z-axis rudder coupling gain and the second z-axis rudder coupling gain represent the amount of rotation of the UAV around the z-axis in the inertial frame, respectively.

[0036] The total drift term is expressed as ;in, Represents the translational state vector The corresponding translational drift term is calculated as follows: ; in, , , These represent the deterministic components of the translational momentum of the UAV along the x, y, and z axes in the inertial frame that can be incorporated into the drift term.

[0037] at the same time, Represents the rotational state vector The corresponding rotational drift term is calculated as follows: ; Through the above construction, this embodiment realizes the explicit injection of thrust triaxial decomposition input and control surface deflection angle contribution to translational aerodynamics under a unified framework, upgrading the system from the traditional underdriven structure of "4 inputs and 6 outputs" to a fully driven affine structure of "6 inputs and 6 outputs", thereby providing a directly callable modeling foundation for the integrated realization of subsequent control allocation, sliding mode control and disturbance observation.

[0038] Besides modeling methods, sliding mode control is widely used due to its strong robustness to uncertainties and external disturbances, and has developed into various forms such as predetermined time or fixed time convergence, incremental sliding mode, and event-triggered adaptive sliding mode. However, most existing methods are still based on the traditional underdriven state-space framework, making it difficult to structurally eliminate nonlinear couplings. Missing terms in the model and coupling residuals are easily included as disturbances, leading to the superposition of control compensation and structural errors, thus limiting the upper limit of accuracy. Especially in the fully driven reversible allocation framework, the virtual quantity calculation, allocation, and compensation chain may structurally amplify the sliding mode switching term, making cross-condition gain matching and chattering suppression more difficult.

[0039] To address this issue, this embodiment proposes a fully driven adaptive sliding mode control method (FA-ADSMC) that combines a dual nonlinear terminal sliding surface with an exponential adaptive gain law.

[0040] Step S2: Based on the desired position and attitude state, construct the position and attitude tracking error, and construct the sliding surface based on the position and attitude tracking error.

[0041] This embodiment uses the desired state of position and attitude. To track the target, a state error is constructed. And introduce the derivative of position and attitude tracking error. Based on this, a double nonlinear power sliding surface is established. To achieve a balance between "rapid convergence of large errors and high-precision steady-state operation with small errors"; among which the sliding surface It consists of linear terms and two sets of nonlinear terms of different powers; sliding surface The expression is: ; ; in, Indicates the position and attitude tracking error. The derivative of the position and attitude tracking error is represented by . Represents a symbolic function. Apply element-wise to vector components. and Let these represent the first and second nonlinear correction orders corresponding to the small and large error intervals, respectively, satisfying... This ensures strong correction capability when far from the equilibrium point, and weak chattering and finite-time approach characteristics when approaching the equilibrium point. Indicates the first An exponential adaptive gain is introduced in this embodiment to adjust the convergence speed and chattering level online at different error stages, so that the gain is adaptively enhanced when the error increases and gradually reduced when the error decreases, while maintaining the necessary robustness margin. Indicates the first The derivative of an exponential adaptive gain, The power exponent of the adaptive law is represented. ; Indicates the first An exponential adaptive gain lower bound is used to avoid insufficient robustness due to gain degradation. Indicates the first An exponential adaptive gain fallback factor is used to suppress unbounded gain growth and reduce chattering risk when the error is small. Indicates the first The growth rate coefficient of the exponential adaptive gain.

[0042] Step S3: Construct the all-drive control structure based on the affine all-drive model and sliding surface to generate the all-drive control input vector.

[0043] This embodiment constructs an all-drive control structure based on an affine all-drive model. Based on this, the all-drive control input vector Designed as a total drive control structure consisting of equivalent control and robust compensation superimposed on the sliding surface. The derivative The stabilization requirements are mapped to the control input space, forming a unified closed-loop constraint on the six-degree-of-freedom state. Equivalent control is used to compensate for known drift terms and reference trajectory dynamic terms, while robust compensation is used to counteract external disturbances, parameter uncertainties, and unmodeled couplings represented by the integrated disturbance term. Furthermore, continuous saturation and hyperbolic tangent injection strategies are employed to reduce high-frequency chattering caused by traditional sign switching while ensuring the approaching velocity.

[0044] The expression for the chatter suppression term is: ; in, This represents the chatter suppression term. Indicates the boundary layer thickness. ; express Time-adaptive robust gain, It is used to adaptively adjust the perturbation suppression strength at different error stages. This represents the hyperbolic tangent function.

[0045] The adaptive robust gain The derivative The expression is: ; in, Indicates the sliding mode threshold. ; Indicates the reference gain. ; , and These represent the first coefficient, the second coefficient, and the third coefficient, respectively. Used to improve suppression capability when the error is large. Used to suppress gain divergence Used to reduce chatter by falling back to the baseline level within a small error range.

[0046] The expression for the all-drive control structure is: ; in, This represents the all-wheel drive control input vector; The second derivative of the desired state representing the position and attitude state. This represents the inverse of the input gain matrix. Used to map the desired acceleration and sliding mode approach requirement to the all-drive input space. This indicates the drift term.

[0047] This embodiment reorganizes the controller structure based on an affine all-drive model to form an all-drive control structure. At the controller level, it performs structured extraction and explicit calculation of aerodynamic effects that are traditionally classified as disturbances in modeling, reducing the risk of oscillations caused by applying unnecessary feedforward gains to unknown terms from the source. At the same time, this embodiment uses the double power terminal configuration of the sliding mode surface body to achieve the unification of "far-end superlinear accelerated convergence" and "near-end continuous vibration suppression and refinement". It also uses exponential adaptive gain to quickly scale the error and disturbance amplitude, enabling rapid gain increase under strong disturbance conditions to accelerate convergence and adaptive gain decrease under weak disturbance conditions to suppress chattering. Thus, under the unified control of the six channels of the all-drive system, it completes online matching and robust compensation for coupled residuals, abrupt changes and slowly varying composite disturbances, effectively overcoming the limitations of traditional fixed gain or single sliding mode structures in all-drive scenarios, such as oscillation amplification, energy scheduling imbalance and weakening of decoupling targets.

[0048] Step S4: Solve for the translational and rotational control quantities based on the all-drive control structure to achieve UAV control.

[0049] This embodiment is based on modular decoupling control quantity solution of the all-drive control structure. To improve the feasibility and anti-coupling numerical stability of online solution, the six-degree-of-freedom affine all-drive model is modularly divided into translational and rotational subsystems, and a block-decoupling control quantity solution strategy is adopted. The all-drive control input vector is... Decomposed into a translational control input vector consisting of translational channel control quantities. and the rotation control input vector composed of rotation channel control quantities , and input gain matrix Written in block format as , , and Let these represent the first sub-gain matrix, the second sub-gain matrix, and the third sub-gain matrix, respectively. The second sub-gain matrix... The gain matrix represents the effect of the rotation channel control quantity on the translational acceleration of the UAV in each axis of the inertial frame, thereby reducing the coupling terms to... The formal representation is explicit, and the numerical amplification caused by direct inversion is reduced by solving in the order of "rotation first, then translation"; in the online implementation, this embodiment uses the rotation control input vector from the previous moment. To construct the coupling approximation term for the translational channel, the translational control input vector is first obtained. Then, the rotation control input vector is obtained from the rotation channel. Ultimately, this achieves independent and adjustable, yet consistent, full-drive decoupled control of the three translational and three rotational motions.

[0050] In this embodiment, the affine all-drive model is decomposed into a translational model and a rotational model. The expressions for the translational model and the rotational model are as follows: ; in, Represents the translational state vector. Represents the rotational state vector. and Let represent the second derivatives of the translational state vector and the rotational state vector, respectively; and These represent the drift terms corresponding to the translational and rotational state vectors, respectively. and These represent the combined disturbance terms corresponding to the translational state vector and the rotational state vector, respectively.

[0051] Translational control input vector and rotation control input vector The expression is: ; Translational control input vector and rotation control input vector The solution result is expressed as: ; ; in, Represents the translational state vector The second derivative of the expected state, Represents the rotational state vector The second derivative of the expected state; The rotation control input vector from the previous moment is used to construct a coupling compensation approximation term to reduce the online solution burden and prevent algebraic loops. The inverse matrix of the first sub-gain matrix; The inverse matrix of the third sub-gain matrix; For the sliding surface of the translational subsystem, For the sliding surface of the rotating subsystem.

[0052] The sliding surface of the translational subsystem is defined as: ; in, This indicates the translational position tracking error. The derivative of the translational position tracking error is expressed. and These represent the orders of the first and second translational nonlinear corrections, respectively; furthermore... This represents the diagonal gain matrix of the low-power error terms in the translational three-channel system. , and These represent the first x-axis translational gain, the first y-axis translational gain, and the first z-axis translational gain, respectively. The diagonal gain matrix represents the high-power error terms of the translational three-channel circuit. , and These represent the second x-axis translational gain, the second y-axis translational gain, and the second z-axis translational gain, respectively.

[0053] Rotating subsystem sliding surface Defined as: ; in, Indicates the rotational attitude tracking error. The derivative of the rotational attitude tracking error; and These represent the orders of the first and second rotational nonlinear corrections, respectively; furthermore... This represents the diagonal gain matrix representing the low-power error term of the three-channel rotation, where... , and These represent the first x-axis angular gain, the first y-axis angular gain, and the first z-axis angular gain, respectively. This represents the diagonal gain matrix of the high-power error terms in the three-channel rotation. , and These represent the second x-axis angle gain, the second y-axis angle gain, and the second z-axis angle gain, respectively.

[0054] This embodiment achieves consistency between the independent solution of the rotation channel and the overall closed loop by solving in blocks.

[0055] Furthermore, disturbance observation and compensation are crucial for the realization of closed-loop control for all-wheel-drive fixed-wing aircraft. All-wheel drive does not simply reduce the impact of disturbances; instead, it makes the disturbances exhibit a more complex structural form: residual terms introduced by model upgrades and variable elimination are redistributed across multiple channels in the form of "structural residuals"; external gusts, aerodynamic parameter drift, and unmodeled coupling are more likely to overlap with the control allocation process, forming a complex disturbance with both abrupt and gradual changes. If the coarse-grained estimation approach of encompassing all disturbances, fixed bandwidth, and fixed gain is still used, problems such as blurred boundaries between interpretable and non-interpretable terms, estimation bias, and noise amplification can easily arise, even weakening the decoupling advantages of all-wheel drive.

[0056] This embodiment addresses the challenges of complex channel distribution, including the superposition of structural residuals and external disturbances, the coexistence of abrupt and gradual changes, and the presence of complex channel distributions in disturbances after all-drive modeling. It proposes an enhanced disturbance observer driven by the Disturbance Structure Response Indicator (DSRI). Leveraging the explicit affine structure and reversible input mapping of the all-drive system, interpretable control-induced aerodynamic terms are extracted from the disturbance, transforming the observed object into a purer channel residual. Furthermore, this embodiment uses a sliding time window to jointly characterize the disturbance's rate of change and energy density, constructing the DSRI for online identification of the disturbance structure. Moreover, this embodiment employs Sigmoid weighting to adaptively schedule the observation gain, achieving both rapid tracking of abrupt disturbances and high-precision estimation of strong steady-state disturbances without relying on a fixed high gain. This solves the estimation lag and noise amplification problems common in traditional extended state observers and conventional sliding mode observers in all-drive scenarios, ensuring the all-drive system's closed-loop decoupling performance and engineering robustness.

[0057] Step S5: Construct an enhanced perturbation observer to observe the comprehensive perturbation term of the affine all-drive model.

[0058] This embodiment provides an enhanced superspiral sliding mode perturbation observation technique driven by the DSRI (Structured Disturbance Response Index) comprehensive perturbation structure response. This embodiment uses an all-wheel-drive fixed-wing UAV as the object, unifying external uncertainties such as gusts, wakes, and unmodeled couplings into the comprehensive perturbation term. The system was rewritten to facilitate enhanced disturbance observer injection and control compensation by constructing auxiliary variables. With observed sliding mode variables This forms a closed-loop interface for "observation compensation," enabling the enhancement of the disturbance compensation term generated by the disturbance observer. It can embed control laws in a way that allows for direct input substitution; to avoid unnecessary high-frequency amplification introduced by traditional adaptive sliding mode disturbance observation in small error intervals and to improve estimation accuracy in scenarios with rapidly changing disturbances, this embodiment adopts a "regression-based" piecewise update strategy: in When the gain is large, the classical gain is increased to accelerate convergence. After entering the small error range, a regression term is introduced to cause the gain to fall back to the lower bound of the reference, thus balancing speed and smoothness.

[0059] To facilitate the design of the enhanced perturbation observer, the position and attitude states are... Record , Represents the generalized pose state vector. A state form written to facilitate the design of enhanced disturbance observers, the expression is: ; in, The derivative of the generalized pose state vector is represented by . Represents the generalized velocity state vector. This represents the integrated control input vector. It represents the derivative of the generalized velocity state vector.

[0060] This embodiment enhances the disturbance observer by designing a more robust disturbance observer. To approximate the result, an auxiliary variable is introduced. Construct a disturbance observation model, the expression of which is: ; in, Representing auxiliary variables The derivative of, This represents the disturbance compensation term that enhances the output of the disturbance observer. Used to approximate the comprehensive disturbance term .

[0061] Define the observed sliding mode variable for: ; in, This represents the observed sliding mode variable, used to characterize the deviation between the actual dynamics and the auxiliary dynamics.

[0062] Observation of sliding mode variables The derivative The expression is: ; when Approaching hour, It can converge within a finite time and suppress the perturbation estimation error.

[0063] Consider the continuous form of the disturbance compensation term as follows: ; in, Indicates a positive coefficient. , Minimal positive numbers are used to avoid singularities. Indicates the superspiral assisted state. The derivative of the superspinning auxiliary state; and These represent the first adaptive gain adjustment and the second adaptive gain adjustment of the enhanced perturbation observer, respectively. The above structure can achieve both fast convergence and chatter suppression under continuous injection conditions.

[0064] Meanwhile, this embodiment modifies the basic gain update stage of the classic adaptive superspiral structure, replacing the original frozen update method in the near-sliding mode area with a continuous update law with a fallback mechanism, expressed as: ; in, This indicates the base gain of the enhanced perturbation observer. This represents the derivative of the base gain used to enhance the perturbation observer; Represents a proportionality constant. ; This represents the gain up-adjustment rate coefficient. ; Represents the regression rate coefficient. ; Indicates the segmentation threshold. ; Indicates the lower bound of the reference. This is used to ensure that the gain falls back in the small error range and increases rapidly in the large error range.

[0065] This embodiment provides a method for constructing a Disturbance Structure Response Synthetic Index (DSRI). Addressing the coexistence of complex disturbances in two dimensions—"disturbance rate of change" and "disturbance energy density"—this embodiment constructs the DSRI to characterize the transient peak intensity and sustained intensity of the disturbance. A smoothing switching function adaptively fuses these two characteristics, providing a unified, structured quantitative basis for observation gain adjustment. The disturbance rate of change is obtained by integrating the disturbance difference within a sliding time window, and the disturbance energy density is constructed from the squared mean of the disturbance within the sliding time window. After normalization, these two values ​​form a dimensionless index, which is used to define the structure response ratio and the switching function. Finally, a DSRI driving signal that updates in real-time with the evolution of the disturbance structure is obtained. The expression for the DSRI driving signal is: ; in, Indicates time, Indicates the length of the sliding time window. The integral time variable within the sliding time window satisfies , This represents the average value of the disturbance estimate within the sliding time window. Indicates a tiny step size. This represents the disturbance estimate obtained by the enhanced disturbance observer. This represents the rate of change of the disturbance, characterizing the degree of rapid change in the disturbance. It represents the energy density of the disturbance, characterizing the sustained energy intensity of the disturbance.

[0066] right and The expression for normalization is: ; in, and These represent the normalized rate of change of the disturbance and the normalized energy density of the disturbance, respectively. Represents the normalized stability factor. To avoid numerical amplification caused by an excessively small denominator, normalization allows the two types of features to be compared and fused on the same scale.

[0067] The expression for the dominance coefficient is: ; in, Indicates a small positive coefficient. To prevent the denominator from being zero, express The dominance coefficient at time is used to measure the dominance of the "rapid change intensity" characterized by the rate of change of the disturbance relative to the "energy intensity" characterized by the energy density of the disturbance.

[0068] Switching function The expression is: ; in, This indicates switching the slope parameter. ; Indicates the discrimination threshold. Indicates the switching function, A continuous and smooth switching is achieved between the two perturbation structures, namely, rapid change-dominated and energy-dominated, ultimately forming the perturbation structure response comprehensive index DSRI, expressed as: ; The Disturbance Structural Response Index (DSRI) integrates the instantaneous rate of change of the disturbance estimate with the energy statistics within the sliding time window, using the normalized amount of the disturbance rate of change. Normalized quantity of perturbation energy density The weighted combination of these factors constructs a real-time "disturbance trend extraction" mechanism that reflects the severity and duration of the disturbance in the time domain. It can reflect the structural characteristics of the disturbance in real time and serve as a driving signal for gain adjustment.

[0069] This embodiment introduces a disturbance compensation term, which enhances the output of the disturbance observer, into the all-drive control structure. This embodiment introduces the disturbance compensation term... Coupled with the Disturbance Structural Response Synthesis Index (DSRI), an enhanced gain regulation law driven by the DSRI is constructed, enabling the observed injection intensity to adaptively increase with the disturbance structure. This achieves an integrated closed-loop estimation, compensation, and control within the same input mapping framework as the all-drive control structure. Specifically, this embodiment superimposes the observation gain from the reference value and the structural response gain, establishing a same-scale coupling relationship between the gains, thereby enhancing the disturbance compensation term output by the disturbance observer. It can directly access the all-drive control input channel to control the comprehensive disturbance term. Online compensation improves estimation accuracy, convergence speed, and engineering robustness under complex perturbation environments.

[0070] First adaptive gain adjustment to enhance the perturbation observer Second adaptive gain adjustment The expression is: ; in, Represents the structural response gain coefficient. ; This represents the coefficient of coupling relationship at the same scale. Used to establish and The same-scale coupling relationship ensures the consistency and adjustability of the superspiral injection structure.

[0071] Disturbance compensation item As a disturbance compensation quantity that can be directly replaced and superimposed as an input, the all-drive control structure is introduced, and the expression of the all-drive control structure is adjusted to: ; in, This represents the all-wheel drive control input vector. This indicates that the disturbance compensation term generated by the enhanced disturbance observer enters the closed loop in a manner that allows for "direct substitution of input." and These represent the position and attitude tracking error and the sliding surface, respectively. express Time-adaptive robust gain, The boundary layer thickness is indicated. In this embodiment, the observation gain adjustment driven by the Disturbance Structure Response Synthesis Index (DSRI) is combined with the full-drive input mapping to achieve unified injection of enhanced disturbance observation compensation under a six-degree-of-freedom full-drive structure, thereby significantly improving the disturbance estimation and control compensation performance in environments with large envelopes, rapid disturbances, and strong coupling.

[0072] To verify the effectiveness and engineering feasibility of the method in this embodiment, a numerical simulation experiment of cyclic trajectory was further conducted. The simulation object adopted a fully driven explicit affine model. To ensure the fairness of the comparison, all comparison algorithms used the same aircraft parameters, the same initial values ​​and trajectory settings, the same disturbance injection form, and the same actuator constraint range. The system, initial values, and trajectory parameters are summarized in Table 1 below.

[0073] The disturbance injection adopts a composite disturbance form (sine term and bias term superposition), and is set according to channel and time segment to form disturbance conditions of sudden change, gradual change and channel heterogeneity: the disturbance parameters of translational channel and rotational channel are shown in Table 2. This setting makes the system face the superposition of aerodynamic coupling residual, external gust, wake and unmodeled terms at different stages of cyclic maneuver, so as to verify the comprehensive advantages of the method in this embodiment in the global scenario.

[0074] The expression for the combined disturbance input is: ; in, Indicates the first The effect on the first time period The combined disturbance input of each channel; and These represent the amplitudes of two sets of sinusoidal disturbance components, used to characterize periodic external disturbances of different intensities; and These represent the angular frequencies of the corresponding sinusoidal perturbation components, used to characterize how fast the perturbation changes; and These represent two sets of oscillatory disturbance terms that change with time; Indicates the first The first stage The constant bias perturbation applied to each channel is used to simulate steady-state aerodynamic deviations, model mismatches, or continuous external bias effects. The perturbation injection takes the form of segmented composite perturbations for each channel: within a given time window, it consists of two sets of sinusoidal terms and bias terms superimposed, with different amplitudes, frequencies, and biases set in different channels to simulate the time-varying nature of external perturbation intensity and structural characteristics during cyclic maneuvers.

[0075] In this embodiment, except for the initial 10-second disturbance which is 0, the remaining disturbances are divided into three time intervals: [10, 20) s, [20, 40) s, and [40, 60] s; the parameter values ​​for each channel in each interval are shown in Table 2. As can be seen from the table, the disturbance amplitude and bias of the translational channel exhibit a step-like change in different time intervals, reflecting abrupt disturbance conditions; the disturbance of the rotational channel emphasizes continuous bias and frequency changes, reflecting a gradually changing disturbance condition. This combination of abrupt, gradually changing, and channel heterogeneous configurations is used to verify the ability of the Disturbance Structure Response Integrated Index (DSRI) of the method in this embodiment to identify the disturbance structure online, as well as the stability and feasibility of the control compensation chain during the coupling enhancement stage.

[0076] To illustrate the simulation scenario and comparison object, this embodiment selects a cyclic trajectory as the three-dimensional reference maneuver task. A comparison of three-dimensional spatial trajectory tracking in cyclic trajectory mode is shown below. Figure 2 For six-channel state tracking in cyclic trajectory mode, see [link / reference]. Figure 3 For a comparison of the absolute errors of the six channels in the cyclic trajectory mode, see [link to comparison]. Figure 4 Comparison of disturbance observation results can be found in [link / reference]. Figure 5 For a comparison of control inputs in cyclic trajectory mode, see [link / reference]. Figure 6 For a comparison of actuator control surface commands in cyclic trajectory mode, see [link to documentation]. Figure 7 The controllers compared in this embodiment include: FA-ADP (all-drive adaptive double power sliding mode in this embodiment), ordinary sliding mode (SMC), super spiral sliding mode (PFSMC), incremental sliding mode (GFSMC), layered double power sliding mode (MFSMC), and fixed gain double power sliding mode (BRSMC).

[0077] The controller is compared with non-singular terminal observers (FTDO), extended sliding mode observers (ESMDO), and robust superspiral observers (LDOB) to demonstrate the comprehensive advantages of the method in this embodiment.

[0078] The accuracy analysis of the cyclic trajectory tracking is explained, and a spatial comparison diagram of the cyclic trajectory reference and the actual trajectory is shown below. Figure 2 ,Depend on Figure 2 It can be determined that the actual trajectory of the method in this embodiment can more closely follow the reference cyclic trajectory, especially in the transition section where curvature changes and attitude adjustments are more frequent, maintaining continuous and smooth tracking without obvious swaying or accumulated deviation. This indicates that the six-channel unified control under the all-drive reversible input mapping can effectively suppress the tracking degradation caused by "strong coupling of translation and rotation". See the six-channel state tracking diagram. Figure 3 ,Depend on Figure 3It can be determined that the method in this embodiment exhibits more consistent convergence behavior in both translational and rotational channels; the tracking curve is close to the reference and the steady-state fluctuation amplitude is smaller, indicating that the dual nonlinear terminal sliding surface can simultaneously achieve rapid far-end correction and continuous near-end vibration suppression and refinement. Furthermore, the exponential adaptive gain can rapidly rescale and improve convergence speed during the disturbance enhancement phase, and adaptively decrease after the disturbance weakens to reduce chattering and noise sensitivity. See the six-channel absolute error comparison chart. Figure 4 , Figure 4 This further demonstrates that the method in this embodiment has a smaller error envelope and faster recovery across the entire time domain. In particular, it can still achieve faster regression during the period of disturbance segment switching and coupling enhancement, reflecting the synergistic effect of structured aerodynamic influence extraction and adaptive sliding surface considering both near and far ends.

[0079] To achieve quantitative evaluation, this embodiment uses full-time domain RMSE as a comprehensive accuracy index. The smaller the RMSE, the higher the tracking accuracy, as shown in Table 3.

[0080] Table 3 can be used for comparison. Figure 4 The error time-domain curves yielded consistent conclusions: the method in this embodiment achieved lower RMSE in most channels and had better overall performance, indicating that it can not only maintain a smaller error in steady state, but also recover faster in stages of increased disturbance or significant channel coupling, thereby maintaining more stable tracking quality in the entire time domain.

[0081] The analysis of disturbance estimation and compensation effects is explained, and a comparison of disturbance observation results is provided (see [link to relevant documentation]). Figure 5 . Figure 5 A comparison curve between the actual and estimated values ​​of the disturbance is presented. The comparison shows that fixed-gain observers typically face a contradiction: small gain results in slow response to abrupt disturbances, while large gain amplifies steady-state noise, manifesting as estimation lag during abrupt changes and high-frequency jitter or noise amplification in the steady-state estimation curve. The method in this embodiment uses a sliding time window to jointly characterize the disturbance's rate of change and energy density, constructing a Disturbance Structure Response Index (DSRI) to achieve online identification of the dominant disturbance structure. Based on this, it drives the enhanced disturbance observer gain to perform Sigmoid-weighted adaptive scheduling: when the disturbance exhibits rapid change dominance, the gain is adaptively increased to accelerate the response; when the disturbance exhibits energy dominance and tends towards a steady state, estimation accuracy is improved and noise amplification is suppressed; when the disturbance weakens, a fallback mechanism is triggered to reduce noise sensitivity. Figure 5 As can be seen, the method in this embodiment can more quickly approximate the actual disturbance in the abrupt change phase, and the estimation is smoother and has smaller deviations in the steady-state phase. This ensures that the compensation chain achieves a better trade-off between speed, accuracy, and noise suppression, and provides... Figure 4 The error fast regression shown provides direct support.

[0082] The analysis of control inputs and actuator constraints is explained below. A comparison of control inputs in cyclic trajectory mode is provided. Figure 6For a comparison of actuator control surface commands, see [link / reference]. Figure 7 . Figure 6 The time-domain curves and variation characteristics of the all-drive control input vector are given. It can be observed that the overall control input of the method in this embodiment is more continuous, and the spikes and high-frequency jitters are weaker. This indicates that the method in this embodiment, through all-drive reversible allocation and controller structure reorganization, structurally extracts and explicitly calculates the aerodynamic effects that are traditionally classified as disturbances. This reduces the "sliding mode switching term structural amplification" induced by applying unnecessary feedforward gain to unknown terms from the source, thereby reducing the risk of chattering in the translational direction. Figure 7 The method in this embodiment displays the changes in control surface commands over time. While satisfying the control surface amplitude constraints, it maintains smoother command changes, with shorter duration and lower frequency of saturation segments. This indicates that the method in this embodiment achieves a more reasonable trade-off between convergence speed, constraint management, and command smoothness, and has better engineering feasibility.

[0083] The specific contents of Tables 1, 2, and 3 are as follows:

[0084]

[0085]

[0086] The analysis of energy loss and robustness indicators is explained. To obtain an overall robustness evaluation of the six channels, this embodiment sets the error vector of the six channels as follows: According to the tolerance standard Dimensionless values ​​are generated and synthesized into a single comprehensive error amplitude. , represented as: ; in, , , , , and They represent , , Three translational channels and , , The permissible error scale parameters for the three rotation channels are taken as follows: For 0.05m, It is 0.1 rad; Indicates the first Each channel at time Tracking error, Indicates the first The permissible error scale corresponding to the channel, and .

[0087] Based on this, two robustness indicators are defined: Excess Duration ; Indicates duration; peak exceedance multiple .

[0088] thus, The smaller The smaller the value, the stronger the overall robustness of the six channels. Furthermore, for ease of standardized sorting, the table provides values ​​based on... and Robustness index obtained by fusion (The larger the value, the stronger the robustness). See Table 4 for specific comparison data.

[0089]

[0090] As shown in Table 4, in the cyclic trajectory task, the method of this embodiment maintains the advantage in robustness index, indicating that its ability to suppress sudden changes, slow-change composite disturbances and channel coupling residuals has cross-trajectory consistency, and can maintain stable closed-loop quality under different maneuver modes.

[0091] To provide a unified quantitative comparison of different control strategies from the perspectives of energy consumption suppression and actuator burden, this embodiment introduces an energy loss proxy index consisting of "translational mechanical positive work" and "rudder surface activity cost".

[0092]

[0093] The integral of positive work of translational machinery is defined as ; The energy consumption cost of control surface activity is defined as .

[0094] in, and These represent the first and second time points, respectively. This represents the transpose of the translational control input vector; Represents the velocity vector; This indicates the deflection angle of each control surface (aileron control surface deflection angle). Elevator surface deflection rudder deflection angle The first derivative of the control surface with respect to time forms the control surface deflection rate vector, which is used to characterize the intensity of control surface activity.

[0095] Comprehensive energy loss index And give the translational average positive power. The above quantities are all cost-based indicators, and the smaller the value, the better. See Table 5 for specific comparison data.

[0096] Combination Figure 6 and Figure 7 It can be seen that the smoother the control input and the less saturation and high-frequency oscillation, the more controllable the energy consumption and the lower the actuator wear and thermal load. Table 5 further verifies from a statistical perspective that the method of this embodiment maintains higher tracking accuracy and stronger robustness while having better or more balanced energy loss indicators, indicating that it achieves a better overall trade-off between accuracy, robustness and energy consumption.

[0097] In summary, this embodiment verifies the synergistic effectiveness of the technical chain of the method under the conditions of cyclic trajectory composite disturbance and actuator constraints: Figures 2 to 4 This shows that the method in this embodiment has higher three-dimensional trajectory tracking accuracy and more consistent six-channel convergence quality in the entire time domain, and the statistical results in Table 3 and Table 2 corroborate each other. Figure 5 This demonstrates that the DSRI-driven enhanced superspiral observation compensation method of this embodiment can achieve adaptive scheduling and fallback between abrupt disturbances and steady-state strong disturbances, alleviating the problems of lag and noise amplification and improving the compensation effect; Figure 6 and Figure 7 This demonstrates that the method in this embodiment provides smoother control input and less saturation under the constraints, resulting in greater engineering feasibility. Tables 4 and 5 further illustrate that the method in this embodiment exhibits more stable robustness under different trajectories and achieves more reasonable energy loss control while maintaining high accuracy and robustness. Therefore, the method in this embodiment can provide a feasible solution integrating modeling, control, and observation compensation for the challenges of fixed-wing underactuated structures, strong coupling, strong time-varying input gain, actuator constraints, and complex disturbances. Equivalent modeling of the full-drive system provides a full-rank reversible drive mapping and supports unified decoupling control of six channels; the double-power sliding mode surface and exponential adaptive gain achieve unified fast convergence and low chattering while suppressing structural amplification; the DSRI (Disturbance Structure Response Index) integrates the disturbance change rate and disturbance energy density to achieve online identification of the disturbance-dominant structure and drive the enhanced disturbance observer's adaptive scheduling and fallback, improving observation speed and accuracy while reducing noise sensitivity. Numerical simulation of cyclic trajectories shows that the method in this embodiment can significantly improve multi-channel tracking accuracy, robustness, and energy consumption suppression under complex disturbance and constraint conditions, demonstrating good engineering application value.

[0098] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling, characterized in that, include: Based on the aerodynamic model of the UAV in the airframe and the coordinate transfer matrix from the airframe to the inertial frame, the gain matrix of the effect of the rotation channel control quantity on the translational acceleration of the UAV in each axis of the inertial frame is determined; the rotation channel control quantity includes the surface deflection angles of the UAV's ailerons, elevators and rudders. The all-drive control input vector is constructed based on the translational and rotational control variables; the translational control variables include the equivalent thrust in each axis of the inertial frame obtained by the comprehensive thrust decomposition of the UAV. A six-degree-of-freedom state vector is constructed based on the position and attitude state of the UAV in the inertial frame; an input gain matrix is ​​constructed between the all-drive control input vector and the six-degree-of-freedom state vector based on the aforementioned gain matrix; An affine all-drive model is constructed based on the all-drive control input vector, the six-degree-of-freedom state vector, and the input gain matrix. Based on the desired position and attitude state, construct the position and attitude tracking error, and construct the sliding surface based on the position and attitude tracking error; An all-drive control structure based on an affine all-drive model and a sliding surface is constructed to generate the all-drive control input vector. Based on the all-drive control structure, the translational and rotational control variables are solved to achieve UAV control.

2. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 1, characterized in that, The expression for the affine all-wheel drive model is: ； in, Indicates the position and attitude state. Represents a six-degree-of-freedom state vector; Indicates the drift term; Represents the input gain matrix; This represents the all-wheel drive control input vector; Indicates the combined disturbance term; , and These represent the translational moments of the UAV along the x, y, and z axes in the inertial frame, respectively. , , These represent the translational accelerations of the UAV along the x, y, and z axes in the inertial frame, respectively. , and These represent the rotations of the UAV around the x-axis, y-axis, and z-axis in the inertial frame, respectively. , , They represent , and The second derivative of , and They represent and The first derivative of; , , These represent the deterministic components related to the translational momentum of the UAV along the x, y, and z axes of the inertial frame, respectively. and The first x-axis rudder coupling gain and the second x-axis rudder coupling gain represent the amount of rotation of the UAV around the x-axis in the inertial frame, respectively. , and The first y-axis rudder coupling gain, the second y-axis rudder coupling gain, and the third y-axis rudder coupling gain represent the amount of rotation of the UAV around the y-axis in the inertial frame, respectively. and The first z-axis rudder coupling gain and the second z-axis rudder coupling gain represent the amount of rotation of the UAV around the z-axis in the inertial frame, respectively. Indicates the quality of the drone; , and These represent the linear contribution vectors of the ailerons, elevator, and rudder of the UAV, respectively. Indicates air dynamic pressure. Indicates the reference area; Represents the coordinate transition matrix from the machine system to the inertial system. The Row vectors; , and These represent the equivalent thrust along the x-axis, y-axis, and z-axis of the inertial frame obtained from the comprehensive thrust decomposition of the UAV, respectively. , and These represent the deflection angles of the ailerons, elevator, and rudder of the UAV, respectively. , , These represent the perturbation terms of the translational motion of the UAV along the x, y, and z axes of the inertial frame, respectively. , , These represent disturbance terms indicating the amount of rotation of the UAV along the x, y, and z axes of the inertial frame, respectively.

3. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 2, characterized in that, , , The translational drift term constituted The expression is: ； ； ； in, Represents the translational state vector The corresponding translational drift term; Indicates air density; Indicates airspeed; Indicates the quality of the drone; , and These represent the x-axis, y-axis, and z-axis reference aerodynamic coefficients obtained by mapping the uncontrollable aerodynamic components of the lift, drag, and lateral force aerodynamic coefficients onto the machine system, respectively. , and These represent the uncontrollable aerodynamic components of the aerodynamic coefficients for lift, drag, and lateral force, respectively. Indicates the angle of attack; , and These represent the roll angular velocities of the UAV in the machine system. Pitch angular velocity With yaw rate dimensional value; , and These represent the basic coefficients for lift, drag, and lateral force, respectively. and They represent the angle of attack. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Indicates the sideslip angle The corresponding lateral force aerodynamic coefficient; and They represent pitch angular velocities, respectively. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Represents roll angular velocity The corresponding lateral force aerodynamic coefficient; Indicates yaw rate The corresponding lateral force aerodynamic coefficient.

4. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 2, characterized in that, The expression for the linear contribution vectors of the ailerons, elevator, and rudder of the UAV is: ； in, Indicates the angle of attack; Indicates the aileron deflection angle The corresponding lateral force aerodynamic coefficient; and These represent the elevator surface deflection angles. The corresponding lift aerodynamic coefficient and drag aerodynamic coefficient; Indicates the rudder surface deflection angle The corresponding lateral force aerodynamic coefficient.

5. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 3, characterized in that, The expression for the sliding surface is: ； ； in, Indicates the sliding surface. Indicates the position and attitude tracking error. The derivative of the position and attitude tracking error; Indicates the first An exponential adaptive gain Indicates the first The derivative of an exponential adaptive gain; Represents a symbolic function. and These represent the first and second nonlinear correction orders, respectively. The power exponent of the adaptive law; Indicates the first The lower bound of the gain for an exponential adaptive gain; Indicates the first A fallback factor for an exponential adaptive gain. Indicates the first The growth rate coefficient of the exponential adaptive gain.

6. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 5, characterized in that, The expression for the all-drive control structure is: ； in, Indicates position and attitude state The second derivative of the expected state; This represents the inverse of the input gain matrix; Indicates the boundary layer thickness; express Adaptive robust gain at time step; This represents the hyperbolic tangent function.

7. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 6, characterized in that, Adaptive Robust Gain The derivative The expression is: ； in, Indicates the sliding mode threshold; Indicates the reference gain; , and These represent the first coefficient, the second coefficient, and the third coefficient, respectively. , and All are greater than 0.

8. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 6, characterized in that, The translational and rotational control variables are solved using a modular decoupled solution method. The solution steps include: All-drive control input vector Decomposed into a translational control input vector consisting of translational channel control quantities. and the rotation control input vector composed of rotation channel control quantities ; Input gain matrix Denoted in block form as , , and Let these represent the first sub-gain matrix, the second sub-gain matrix, and the third sub-gain matrix, respectively. The second sub-gain matrix... The gain matrix represents the effect of the rotation channel control quantity on the translational acceleration of the UAV in each axis of the inertial frame. The affine all-wheel drive model is decomposed into a translational model and a rotational model. The expressions for the translational and rotational models are as follows: ； in, The translational state vector, Let be the rotational state vector. and Let represent the second derivatives of the translational state vector and the rotational state vector, respectively; and These are the drift terms corresponding to the translational and rotational state vectors, respectively. and These are the combined disturbance terms corresponding to the translational state vector and the rotational state vector, respectively; Translational control input vector and rotation control input vector The solution result is expressed as: ； ； in, Represents the translational state vector The second derivative of the expected state, Represents the rotational state vector The second derivative of the expected state; This represents the rotation control input vector at the previous moment. The inverse matrix of the first sub-gain matrix; The inverse matrix of the third sub-gain matrix; This represents the sliding surface of the translational subsystem. Indicates the sliding surface of the rotating subsystem; The expression for the sliding surface of the translational subsystem is: in, This indicates the translational position tracking error. The derivative of the translational position tracking error; and These represent the first and second translational nonlinear correction orders, respectively. Represents a symbolic function; The diagonal gain matrix represents the low-power error term of the translational three-channel circuit. , and These represent the first x-axis translational gain, the first y-axis translational gain, and the first z-axis translational gain, respectively. The diagonal gain matrix represents the high-power error terms of the translational three-channel circuit. , and These represent the second x-axis translational gain, the second y-axis translational gain, and the second z-axis translational gain, respectively. The expression for the sliding surface of the rotating subsystem is: in, Indicates the rotational attitude tracking error. The derivative of the rotational attitude tracking error; and These represent the first and second orders of rotational nonlinear correction, respectively. Let be the diagonal gain matrix of the three-channel low-power error terms, where , and These represent the first x-axis angle gain, the first y-axis angle gain, and the first z-axis angle gain, respectively. To rotate the diagonal gain matrix of the high-power error terms of the three channels, , and These represent the second x-axis angle gain, the second y-axis angle gain, and the second z-axis angle gain, respectively. Represents a symbolic function.

9. The six-degree-of-freedom sliding mode control method for fixed-wing UAVs based on all-drive modeling as described in claim 6, characterized in that, The combined perturbation term of the affine all-drive model is observed using an enhanced perturbation observer, the expression of which is: ； ； ； ； ； ； in, Indicates time, This represents the output of the enhanced disturbance observer used to approximate the integrated disturbance term. Disturbance compensation item, and These represent the first adaptive gain adjustment and the second adaptive gain adjustment of the enhanced disturbance observer, respectively. Represents the observed sliding mode variable; This indicates the base gain for enhancing the perturbation observer; and These represent the normalized rate of change of the disturbance and the normalized energy density of the disturbance, respectively. Indicates a superspiral assisted state; The derivative of the superspinning auxiliary state; Indicates a positive coefficient; This indicates the base gain of the enhanced perturbation observer. This represents the comprehensive index of the response of the disturbed structure. Represents the structural response gain coefficient. This represents the coefficient of coupling relationship at the same scale. Indicates the switching function; This indicates the corresponding position and attitude state. The generalized pose state vector, The derivative of the generalized pose state vector is represented by . Represents the generalized velocity state vector. This represents the integrated control input vector. The derivative of the generalized velocity state vector is given by: As an auxiliary variable, The derivative of the auxiliary variable; After introducing the disturbance compensation term from the enhanced disturbance observer output into the all-drive control structure, the expression of the all-drive control structure is adjusted to: 。 10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which is used to execute the six-degree-of-freedom sliding mode control method for a fixed-wing UAV based on all-drive modeling as described in any one of claims 1-9.

Images