A flying-wing unmanned aerial vehicle dynamic inverse glide control method based on cascade observation

Abstract

The application belongs to the technical field of flying wing unmanned aerial vehicle control, and discloses a flying wing unmanned aerial vehicle dynamic inverse glide control method based on cascade observation. First, a flying wing unmanned aerial vehicle dynamics model is established, a nonlinear dynamic inverse method is adopted to design attitude loop and angular velocity loop control law respectively, and unmanned aerial vehicle three-channel dynamic decoupling control is realized. In view of the problem that traditional dynamic inverse control is sensitive to model accuracy and weak in anti-interference ability, a cascade extended state observer is introduced to realize real-time estimation and dynamic compensation of model uncertainty and wind disturbance. Meanwhile, a control distribution strategy based on weighted pseudo-inverse is designed to realize high-precision distribution of control moment. The simulation result shows that the method can effectively inhibit the three-channel coupling effect in the glide process of the unmanned aerial vehicle, significantly improve the trajectory tracking accuracy, attitude stability and system robustness, and adapt to the high-precision glide control requirement of the flying wing unmanned aerial vehicle.

Description

Technical Field

[0001] This invention belongs to the field of flying-wing unmanned aerial vehicle (UAV) control technology, and relates to a dynamic reverse glide control method for flying-wing UAVs based on cascaded observation. Background Technology

[0002] Flying-wing UAVs, with their high lift-to-drag ratio and good stealth performance, can better meet the comprehensive requirements for range and stealth capabilities, making them one of the important configurations for future UAV development. Compared with conventional UAVs, flying-wing UAVs eliminate the traditional tail, resulting in more complex longitudinal, lateral, and directional stability characteristics, and significant coupling between multiple control surfaces. Especially during the glide phase, it is necessary not only to accurately track the predetermined glide trajectory but also to cope with various influences such as wind disturbances and model uncertainties. Therefore, the glide control of flying-wing UAVs is not only a high-precision trajectory tracking problem under strong disturbance conditions but also a nonlinear cooperative control problem under strongly coupled, multi-control-surface constraints.

[0003] Due to the lack of a vertical tail and traditional tail fin stabilizing surface, the longitudinal and directional dynamic characteristics of flying-wing UAVs are generally weaker than those of conventionally laid-out aircraft. They are more prone to interactions between roll, yaw, and sideslip under crosswind and high angle-of-attack approach conditions. While published research has yielded significant results regarding the glide path control of UAVs, laying the foundation for such control, much attention remains focused on longitudinal glide path tracking. Research on the channel interactions during low-speed approaches is still insufficient. Because of their inherent characteristics, flying-wing UAVs cannot simply apply the longitudinal control framework of conventionally laid-out aircraft; instead, a redesign of the control law from a three-channel decoupling perspective is necessary.

[0004] In addition to self-coupling, the descent process is also affected by strong disturbances and model uncertainties. Zhen et al. proposed a control method based on preview control and particle filtering, which improved the control accuracy under dynamic constraints (Zhen Z, Jiang S, Ma K. Automatic carrier landing control for unmanned aerial vehicles based on preview control and particle filtering[J]. Aerospace Science and Technology, 2018, 81: 99-107.). Zhen et al. proposed a multivariable adaptive control scheme to improve the system's adaptability to model uncertainties and external disturbances (Zhen Z, Tao G, Yu C, et al. A multivariable adaptive control scheme for automatic carrier landing of UAV[J]. Aerospace Science and Technology, 2019, 92: 714-721.). Lee et al. employed sliding mode guidance and control to predict and compensate for reference trajectory deviations during terminal approach, thereby enhancing the robustness of the control system (Lee S, Lee J, Lee S, et al. Sliding Mode Guidance and Control for UAV Carrier Landing[J]. IEEE Transactions on Aerospace and Electronic Systems, 2019, 55(2): 951-966.). Lungu et al. combined backstepping, sliding mode control, and extended state observer methods to construct a control system (Lungu M, Chen M, Vîlcic D A. Backstepping and Sliding Mode-Based Automatic CarrierLanding System with Deck Motion Estimation and Compensation[J]. Aerospace,2022, 9(11): 644.). These studies show that strong disturbances and model uncertainties are factors that cannot be ignored in glide slope control. Therefore, it is necessary to introduce disturbance estimation and compensation mechanisms into the control framework to reduce the dependence of the control law on an accurate model.

[0005] Finally, flying wing UAVs typically employ composite control surfaces such as elevons, drag rudders, pitch flaps, and all-moving wingtips. The deflection of a single control surface often generates torques in multiple directions simultaneously. For example, although drag rudders have strong heading control capabilities, their torque coupling is severe and they need to be used in conjunction with other control surfaces. Therefore, it is also necessary to design a control distribution method for three channels to achieve glide path tracking. Summary of the Invention

[0006] To address the issues of strong coupling and strong interference during the descent of flying-wing UAVs, this invention proposes a dynamic inverse descent control method based on cascaded observation. First, a dynamic model of the flying-wing UAV is established. Then, the attitude loop and angular velocity loop control laws are designed using the nonlinear dynamic inverse (NDI) method to achieve dynamic decoupling control of the UAV across three channels. To address the sensitivity of traditional dynamic inverse control to model accuracy and its weak anti-interference capability, a cascaded extended state observer (CESO) is introduced to estimate and dynamically compensate for model uncertainties and wind disturbances in real time. Simultaneously, a control allocation strategy based on weighted pseudo-inverse is designed to achieve high-precision allocation of control torque. Simulation results show that this method can effectively suppress the three-channel coupling effect during the UAV's descent, significantly improve trajectory tracking accuracy, attitude stability, and system robustness, and meet the high-precision descent control requirements of flying-wing UAVs.

[0007] The technical solution of the present invention: A dynamic inverse glide path control method for a flying-wing unmanned aerial vehicle (UAV) based on cascaded observations includes dynamic modeling of the flying-wing UAV, design of the trajectory loop guidance law, dynamic inverse control based on cascaded extended state observations, and control allocation based on weighted pseudo-inverses. The specific steps are as follows: Step (1) Dynamic modeling of flying-wing UAV The force and moment coefficients of the established flying-wing UAV dynamic model are shown below: (1) (2) In the formula: These are the body coordinate system x Directional force coefficient, y Directional force coefficient and z Directional force coefficient; These are the roll moment coefficient, pitch moment coefficient, and yaw moment coefficient, respectively. For the angle of attack, Sideslip angle, For speed, These are roll rate, pitch rate, and yaw rate, respectively. For control surface configuration, For thrust, The left inner leading edge flap and the right inner leading edge flap are the two types of flaps. The outer leading edge flaps are on the left and right sides. The left and right all-moving wingtips are respectively. The left and right elevons are the left and right elevons. The left and right pitch flaps are the pitch flaps. These are the left-side drag rudder and the right-side drag rudder.

[0008] Based on the above expressions for force and moment coefficients, the expressions for force and moment of the flying-wing UAV are further obtained as follows: (3) In the formula: , , These are the body coordinate systems. x Directional force, y Directional force and z Directional force, , , These are respectively the rolling moment, pitching moment, and yaw moment. Atmospheric density; Wing area; For the average aerodynamic chord length, For wingspan, It is the thrust vector arm. The pitch thrust vector This is the yaw thrust vector.

[0009] The dynamic model of the flying-wing UAV is shown below: (4) In the formula: superscript " " represents the first derivative; For quality; It is the acceleration due to gravity; They are respectively x,y,z Three-axis velocity components; These are the first derivatives of the roll angle, pitch angle, and yaw angle, respectively. These are roll angle, pitch angle, and yaw angle, respectively. for x,y,z Three-axis speed; It is a constant related to the moment of inertia.

[0010] Step (2) Design of the guiding law for the trajectory loop To achieve trajectory descent control of a flying-wing UAV, it is first necessary to set the descent line and derive attitude loop control commands based on the trajectory loop guidance law.

[0011] The control objectives for the descent phase mainly include longitudinal height tracking and lateral position convergence. The height error and lateral position error are defined as follows: (5) In the formula: This is the height command corresponding to the dynamic glide path; This refers to the actual height. This is for height error; For reference lateral position, This is the actual lateral position. This represents the lateral position error.

[0012] Pitch and roll commands are generated from altitude and lateral position errors, respectively. (6) (7) In the formula: To balance the pitch angle; To control the gain; Indicates the instruction limiting function; This is the reference command for the roll angle; The reference command for the pitch angle; For the time of descent; This is a time parameter.

[0013] The guiding law of the trajectory loop transforms the trajectory tracking problem into an attitude tracking problem. Altitude error is addressed through... Adjusting the vertical downward trend, the lateral error is passed through Adjusting the lateral correction trend provides a reference signal for subsequent inner-loop control design.

[0014] Step (3) Dynamic inverse control based on cascaded extended state observation Based on the attitude control command obtained in step (2), nonlinear dynamic inverse control of the attitude angle is first implemented, and then a cascaded extended state observer is introduced to realize interference observation compensation. The following is a description of each channel.

[0015] (3.1) Design of slow loop attitude angle control law This invention selects the slow loop variable as The slow loop corresponds to the attitude angle loop, from which the following set of nonlinear differential equations can be obtained: (8) In the formula: For the lateral force of the wind axis; For engine thrust; The expression is as follows: (9) Attitude angular velocity vector is Considering the disturbances and model uncertainties during the descent, the uncertainties in the attitude angle loop are uniformly represented as equivalent disturbances. ,get: (10) in (11) (12) To improve the attitude angle loop's ability to compensate for external disturbances and model uncertainties, a cascaded extended state observer is introduced into the attitude angle loop. The virtual input of the attitude angle loop is defined. for (13) It can then be written as (14) Based on the structure of the cascaded extended state observer, the first-stage observer of the attitude angle loop is designed as follows: (15) The second-stage observer is designed as (16) In the formula: This refers to the attitude angle observation error; This is the estimated attitude angle value; Estimating disturbances for the first-level observer; Estimate the remaining perturbation for the second-stage observer; All are observer parameters. The total perturbation estimate for the attitude angle loop is: (17) in: Used to replace attitude angle measurements in error feedback Used to compensate for equivalent disturbances in the attitude angle loop. Observer parameters are calculated using the following formula. (18) In the formula: This represents the bandwidth of the attitude angle loop observer. Let be a third-order identity matrix. Equation (10) can be rewritten as follows: (19) In the formula: For the control parameters, we have the following expression. (20) when and Reversible, therefore the control law for the slow loop of the system is: (twenty one) Therefore, with the introduction of cascaded observers, the attitude angle loop no longer directly uses measured values. Instead of constructing errors, observations are used. Construction error, and utilize Compensation is provided for the equivalent perturbation of the attitude loop.

[0016] (3.2) Design of fast-loop attitude angular velocity control law The fast loop corresponds to the torque equation of the UAV, and the attitude angular velocity loop. As a control input quantity As a control output, it includes all control surfaces. Considering aerodynamic moment model errors, control surface performance deviations, and wind disturbances, the uncertainties in the angular velocity loop are uniformly represented as equivalent disturbances. The fast-loop angular rate equation can be rearranged to obtain... (twenty two) make In the formula (twenty three) (twenty four) Define angular velocity loop virtual input for (25) It can then be written as (26) A cascaded extended state observer is introduced into the angular velocity loop. The first-stage observer is designed as follows: (27) The second-stage observer is designed as (28) In the formula: This refers to the angular velocity error. This is an estimate of the angular velocity; Estimating disturbances for the first-level observer; Estimate the remaining perturbation for the second-stage observer; All are observer parameters. The total perturbation estimate for the angular velocity loop is: (29) in: Used to replace angular velocity measurements in error feedback; Used to compensate for equivalent disturbances in the angular velocity loop. The parameters of the angular velocity loop observer are selected as follows: (30) In the formula: Let be the bandwidth of the angular velocity loop observer. According to formula (22), its calculation formula is as follows: (31) (32) In the formula: For the control parameters, we have the following expression. (33) The control law of the fast loop of the system is (34) The corresponding torque command is thus obtained. Compared to the original dynamic inverse control law for angular velocity, the angular velocity loop, after the introduction of the cascaded observer, passes through... Provide status feedback and through Compensation is provided for errors in the aerodynamic moment model, changes in control surface effectiveness, and the effects of wind disturbance.

[0017] (3.3) Dynamic compensation loop control design During the descent of a flying wing UAV, maintaining speed stability is crucial. Therefore, an automatic throttle module is required to maintain speed. This invention does not employ thrust vectoring; instead, the thrust vector angles are all set to 0, and a dynamic inverse method is used in the design.

[0018] The differential formula for velocity is shown below: (35) In the formula: The drag force is the wind axis resistance. After rearranging and deforming it, we obtain: (36) in (37) (38) In the formula: This represents the thrust corresponding to the current nominal throttle position. The partial derivative of thrust with respect to throttle command; This is the throttle command.

[0019] Define the virtual input of the speed loop for (39) It can be written as (40) A cascaded extended state observer is introduced into the velocity loop. The first-stage observer is designed as follows: (41) The second-stage observer is designed as (42) In the formula: For speed error; This is a speed estimate; Estimating perturbations for the first-level ESO; Estimate the remaining perturbations for the second-level ESO; All are observer parameters. The total perturbation estimate for the angular velocity loop is: (43) in: Used to replace speed measurements in error feedback; Used to compensate for equivalent disturbances in the velocity loop. The velocity loop observer parameters are selected as follows: (44) In the formula: This represents the bandwidth of the velocity loop observer.

[0020] After introducing the cascaded observer, equation (35) is rewritten as follows: (45) In the formula: For the desired approach speed, This is for controlling the speed channel gain. Ultimately, the speed control law is: (46) This method can reduce the impact of wind disturbance and thrust response error on speed maintenance performance, thereby improving the stability of glide path tracking.

[0021] Step (4) Control allocation based on weighted pseudo-inverse The three-axis torque command obtained in step (3) is assigned to all control surfaces of the UAV to achieve control assignment. Control assignment is used to solve the coordinated control problem of multiple control surfaces. According to the desired command and control objective, the control quantity of each control surface is solved. Since the assignment efficiency of the general pseudo-inverse method is low and different control surfaces have different rate limits and bandwidths, different weights should be set for different control surfaces in the actual control law design.

[0022] Let the virtual control quantity of the system be (47) (48) The control effectiveness matrix is ​​denoted as Then the control allocation relationship can be written as (49) Considering that different control surfaces have different travel limits, speed limits, and usage priorities, a weighted pseudo-inverse control allocation is adopted. A performance index function is defined. for (50) The constraints are (51) Then its optimal solution can be written as (52) in (53) In the formula: Assign a matrix to the weighted pseudo-inverse control; This is the rudder surface weight matrix. Through appropriate selection... The allocation results for each control surface are optimized. This allocation method can meet the desired torque requirements of the three axes while taking into account the coordinated control of multiple control surfaces and the actual execution constraints.

[0023] The beneficial effects of this invention are: This invention addresses the core issues of strong coupling and strong interference during the descent of flying-wing UAVs by proposing a dynamic inverse flying-wing UAV control method based on cascaded extended state observation. First, a dynamic model of the flying-wing UAV is established. Then, a nonlinear dynamic inverse method is used to design attitude loop and angular velocity loop control laws respectively, achieving dynamic decoupling control of the UAV across three channels. To address the problems of traditional dynamic inverse control being sensitive to model accuracy and having weak anti-interference capabilities, a cascaded extended state observer is introduced to estimate and dynamically compensate for model uncertainties and wind disturbances in real time. Simultaneously, a control allocation strategy based on weighted pseudo-inverse is designed to achieve high-precision allocation of control torque. Simulation results show that this method can effectively suppress the three-channel coupling effect during the UAV's descent, significantly improve trajectory tracking accuracy, attitude stability, and system robustness, and meet the high-precision descent control requirements of flying-wing UAVs. Attached Figure Description

[0024] Figure 1 This is a block diagram of a dynamic reverse glide control method for flying-wing UAVs based on cascaded observation; Figure 2 This is the control surface configuration for the ICE flying wing UAV; Figure 3 This is a block diagram of a cascaded extended state observer; Figure 4 This is a comparison chart of traditional control methods (PID) and dynamic inverse control methods (NDI); Figure 5 This is a lateral position comparison diagram of the traditional control method (PID) and the dynamic inverse control method (NDI); Figure 6 This is a speed comparison chart between the traditional control method (PID) and the dynamic inverse control method (NDI); Figure 7 This is a comparison chart of angle of attack between the traditional control method (PID) and the dynamic inverse control method (NDI). Figure 8 This is a comparison chart of pitch angles between the traditional control method (PID) and the dynamic inverse control method (NDI). Figure 9 This is a comparison chart of roll angles for traditional control methods (PID) and dynamic inverse control methods (NDI). Figure 10 This is a comparison chart of sideslip angles for traditional control methods (PID) and dynamic inverse control methods (NDI). Figure 11 This is a comparison chart of the trajectory tilt angles of the traditional control method (PID) and the dynamic inverse control method (NDI). Figure 12 This is a comparison chart of the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 13 This is a comparison diagram of the lateral positions of the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 14 This is a speed comparison chart of the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 15 This is a comparison chart of angle of attack between the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 16 This is a comparison chart of pitch angles between the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 17 This is a comparison chart of roll angles between the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 18 This is a comparison chart of the sideslip angles of the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 19 This is a comparison chart of the track inclination angles of the Dynamic Inverse Control (NDI) method and the Cascaded Observation Dynamic Inverse Control (NDI+CESO) method; Figure 20 This is a comparison chart of torque accuracy based on the weighted pseudo-inverse control allocation method; Figure 21 It is a control surface response diagram based on the weighted pseudo-inverse control allocation method. Detailed Implementation

[0025] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0026] The overall flowchart of the dynamic reverse glide control method for flying-wing UAVs based on cascaded observation is as follows: Figure 1 As shown, first according to Figure 2 Dynamic modeling was performed on the control surface configuration of the flying wing UAV shown, and then the design was carried out according to... Figure 3 A cascaded extended state observer is introduced into the dynamic inverse control framework for control design. Figure 4-11 This is a comparison chart of traditional control methods and dynamic inverse control methods. Figure 12-19 This is a comparison chart of the dynamic inverse control method and the cascaded observation dynamic inverse control method. Figure 20-21 The accuracy diagram is shown for the control allocation method based on weighted pseudo-inverse. The specific steps are as follows: (1) Dynamic modeling of flying-wing UAVs The force and moment coefficients of the established flying-wing UAV dynamic model are shown below: (54) (55) In the formula: These are the body coordinate system x Directional force coefficient, y Directional force coefficient and z Directional force coefficient; These are the roll moment coefficient, pitch moment coefficient, and yaw moment coefficient, respectively. For the angle of attack, Sideslip angle, For speed, These are roll rate, pitch rate, and yaw rate, respectively. For control surface configuration, For thrust, The left inner leading edge flap and the right inner leading edge flap are the two flaps. The outer leading edge flaps are on the left and right sides. The left and right all-moving wingtips are respectively. The left and right elevons are the left and right elevons. The left and right pitch flaps are the pitch flaps. These are the left-side drag rudder and the right-side drag rudder.

[0027] Based on the above expressions for force and moment coefficients, the expressions for force and moment of the flying-wing UAV can be further obtained.

[0028] (56) In the formula: , , These are the body coordinate systems. x Directional force, y Directional force and z Directional force, , , These are respectively the rolling moment, pitching moment, and yaw moment. Atmospheric density; Wing area; For the average aerodynamic chord length, For wingspan, It is the thrust vector arm. The pitch thrust vector, This is the yaw thrust vector.

[0029] The dynamic model of the flying wing UAV is shown below: (57) In the formula: For quality; It is the acceleration due to gravity; They are respectively x,y,z Three-axis velocity components; These are the first derivatives of the roll angle, pitch angle, and yaw angle, respectively. These are roll angle, pitch angle, and yaw angle, respectively. for x,y,z Three-axis speed; It is a constant related to the moment of inertia.

[0030] (2) Design of the guiding law for the trajectory loop The control objectives for the descent phase mainly include longitudinal height tracking and lateral position convergence. The height error and lateral position error are defined as follows: (58) In the formula: This is the height command corresponding to the dynamic glide path; This refers to the actual height. This is for height error; For reference lateral position, This is the actual lateral position. This represents the lateral position error.

[0031] Pitch and roll commands are generated from altitude and lateral position errors, respectively. (59) (60) In the formula: To balance the pitch angle; To control the gain; Indicates the instruction limiting function; This is the reference command for the roll angle; The reference command for the pitch angle; For the time of descent; This is a time parameter.

[0032] (3) Design of traditional glide control for flying wing UAVs (3.1) Longitudinal control design Define pitch angle error and velocity error as follows: (61) The longitudinal attitude control law and the velocity control law are written as follows: (62) In the formula: For elevator / elevator commands; This is the throttle command; It is the pitch angular velocity; To control the gain.

[0033] (3.2) Lateral control design In addition to maintaining longitudinal height and speed control, lateral deviations also need to be corrected. Lateral position error, roll angle error, and sideslip angle error are defined as follows: (63) The roll control law is written as (64) The sideslip control law is written as (65) In the formula: , , , These are the commands for all-moving wingtips, drag rudder, elevons, and throttle control. To control the gain.

[0034] Among them, those applied to longitudinal control For unidirectional deflection, used for lateral control. , Differential control. Classical control methods have a clear structure and are simple to implement, but they do not consider three-axis coupling, changes in rudder effect state, and other factors. Coupled with the inter-control operation of pitch control surfaces.

[0035] (4) Dynamic inverse control based on cascaded extended state observation (4.1) Design of slow loop attitude angle control law This invention selects the slow loop variable as For the corresponding attitude angle loop, the following set of nonlinear differential equations can be obtained: (66) In the formula: This refers to the lateral force on the wind axis; For engine thrust; The expression is shown below.

[0036] (67) Attitude angular velocity vector is Considering the disturbances and model uncertainties during the descent, the uncertainties in the attitude angle loop are uniformly represented as equivalent disturbances. The following form is obtained. (68) in (69) (70) To improve the attitude angle loop's ability to compensate for external disturbances and model uncertainties, a cascaded extended state observer is introduced into the attitude angle loop. The virtual input of the attitude angle loop is defined. for (71) It can then be written as (72) Based on the structure of the cascaded extended state observer, the first-stage observer of the attitude angle loop is designed as follows: (73) The second-stage observer is designed as (74) In the formula: This refers to the attitude angle observation error; This is the estimated attitude angle value; Estimating disturbances for the first-level observer; Estimate the remaining perturbation for the second-stage observer; All are observer parameters. The total perturbation estimate for the attitude angle loop is: (75) in: Used to replace attitude angle measurements in error feedback Used to compensate for equivalent disturbances in the attitude angle loop. Observer parameters are calculated using the following formula. (76) In the formula: This represents the bandwidth of the attitude angle loop observer. Let be a third-order identity matrix. Equation (68) can be rewritten as follows: (77) (78) In the formula: For the control parameters, we have the following expression. (79) when and Reversible, therefore the control law for the slow loop of the system is: (80) Therefore, with the introduction of cascaded observers, the attitude angle loop no longer directly uses measured values. Instead of constructing errors, observations are used. Construction error, and utilize Compensation is provided for the equivalent perturbation of the attitude loop.

[0037] (4.2) Design of fast-loop attitude angular velocity control law The fast loop corresponds to the torque equation of the UAV, and the attitude angular velocity loop. As a control input quantity As a control output, it includes all control surfaces. Considering aerodynamic moment model errors, control surface performance deviations, and wind disturbances, the uncertainties in the angular velocity loop are uniformly represented as equivalent disturbances. The fast-loop angular rate equation can be rearranged to obtain... (81) make In the formula (82) (83) Define the virtual input of the angular velocity loop. for (84) It can then be written as (85) A cascaded extended state observer is introduced into the angular velocity loop. The first-stage observer is designed as follows: (86) The second-stage observer is designed as (87) In the formula: This refers to the angular velocity error. This is an estimate of the angular velocity; Estimating disturbances for the first-level observer; Estimate the remaining perturbation for the second-stage observer; All are observer parameters. The total perturbation estimate for the angular velocity loop is: (88) in: Used to replace angular velocity measurements in error feedback; Used to compensate for equivalent disturbances in the angular velocity loop. The parameters of the angular velocity loop observer are selected as follows: (89) In the formula: Let be the bandwidth of the angular velocity loop observer. According to formula (81), its calculation formula is as follows: (90) (91) In the formula: For the control parameters, we have the following expression. (92) The control law of the fast loop of the system is (93) The corresponding torque command is thus obtained. Compared to the original dynamic inverse control law for angular velocity, the angular velocity loop, after the introduction of the cascaded observer, passes through... Provide status feedback and through Compensation is provided for errors in the aerodynamic moment model, changes in control surface effectiveness, and the effects of wind disturbance.

[0038] (4.3) Dynamic compensation loop control design During the descent of a flying wing UAV, maintaining speed stability is crucial. Therefore, an automatic throttle module is required to maintain speed. This invention does not employ thrust vectoring; instead, the thrust vector angles are all set to 0, and a dynamic inverse method is used in the design.

[0039] The differential formula for velocity is shown below: (94) In the formula: The drag force is the wind axis resistance. After rearranging and deforming it, we obtain: (95) in (96) (97) In the formula: This represents the thrust corresponding to the current nominal throttle position. The partial derivative of thrust with respect to throttle command; This is the throttle command.

[0040] Define the virtual input of the speed loop for (98) It can be written as (99) A cascaded extended state observer is introduced into the velocity loop. The first-stage observer is designed as follows: (100) The second-stage observer is designed as (101) In the formula: For speed error; This is a speed estimate; Estimating perturbations for the first-level ESO; Estimate the remaining perturbations for the second-level ESO; All are observer parameters. The total perturbation estimate for the angular velocity loop is: in: Used to replace speed measurements in error feedback; Used to compensate for equivalent disturbances in the velocity loop. The velocity loop observer parameters are selected as follows: (102) In the formula: This represents the bandwidth of the velocity loop observer.

[0041] After introducing the cascade observer, equation (94) is rewritten as follows: (103) In the formula: For the desired approach speed, This is for controlling the speed channel gain. Ultimately, the speed control law is: (104) This method can reduce the impact of wind disturbance and thrust response error on speed maintenance performance, thereby improving the stability of glide path tracking.

[0042] (5) Control allocation based on weighted pseudo-inverse Control allocation algorithms are used to solve coordinated control problems involving multiple control surfaces. Based on the desired command and control objective, they solve for the control variables of each control surface. Since the allocation efficiency of general pseudo-inverse methods is low, and different control surfaces have different rate limits and bandwidths, different weights should be assigned to different control surfaces in practical control law design. Therefore, a weighted pseudo-inverse algorithm can be used for allocation.

[0043] Control allocation is used to solve for the control surface deflection command under the combined action of multiple control surfaces. Let the virtual control quantity of the system be... (105) The control input vector is (106) The control effectiveness matrix is ​​denoted as Then the control allocation relationship can be written as (107) Considering that different control surfaces have different travel limits, speed limits, and usage priorities, a weighted pseudo-inverse control allocation is adopted. The performance index function is defined as follows: (108) The constraints are (109) Then its optimal solution can be written as (110) (111) In the formula: Assign a matrix to the weighted pseudo-inverse control; This is the control surface weight matrix. This allocation method can meet the desired torque requirements of the three axes while taking into account the coordinated control of multiple control surfaces and the actual execution constraints.

[0044] (6) Simulation condition design The design of the real-time flight altitude of the drone is shown in the following formula.

[0045] (112) In the formula: This is the initial descent height; For flight time; For altitude reference instructions; The trajectory tilt angle is the reference command. Table 1 shows the simulation conditions for glide slope control.

[0046] Table 1 Initial Inputs for Simulation

[0047] The simulation verification selected a trim speed of 80 m / s, an angle of attack of 14.32°, and a reference track inclination of -3.83° to track a specified altitude. The lateral displacement deviation was set to 20 m, and the simulation time was 90 s. Random wind interference was introduced in the last 20 s.

[0048] (7) Simulation comparison between traditional control method and dynamic inverse control method Figure 4-11 A comparison of the responses of the traditional control method and the nonlinear dynamic inverse control method under the same operating conditions is presented. It can be seen that both methods can complete the glide path tracking, but the nonlinear dynamic inverse method performs better in terms of transient response and steady-state accuracy. In the altitude channel, the root mean square error of altitude tracking by the nonlinear dynamic inverse control method decreased from 4.92m to 4.24m, indicating that its tracking of the glide path is more accurate. In the lateral slip channel, the nonlinear dynamic inverse method can track the 20m lateral command faster, the lateral position overshoot decreased from 5.09% of the traditional method to 0.12%, and the settling time shortened from 49.69s to 14.06s, indicating that its lateral correction process is smoother.

[0049] In terms of velocity and attitude response, the nonlinear dynamic inverse method can maintain the velocity around 80 m / s, with a maximum velocity deviation of 2.29 m / s and a steady-state fluctuation amplitude of 0.52 m / s, both smaller than those of the traditional method. In the responses to angle of attack, pitch, and track tilt, the overshoot and steady-state oscillations of the nonlinear dynamic inverse method are reduced, indicating improved dynamic quality of the longitudinal channel. Regarding lateral coupling, the traditional method causes a large sideslip angle response during lateral correction, with a peak sideslip angle of 3.27°, while the corresponding peak value of the nonlinear dynamic inverse method is reduced to 2.65°. This demonstrates that nonlinear dynamic inverse control can effectively reduce the interference of lateral position correction on the attitude channel, improving the decoupling control capability and attitude stability of the flying wing UAV during descent.

[0050] (8) Comparative Simulation of Dynamic Inverse Control Method and Cascaded Observation Dynamic Inverse Control Method Due to the strong model dependence of the dynamic inverse control method, this invention introduces a cascaded extended state observer. The simulation process considers model uncertainty and performs a pull test to compare its robustness and accuracy. In the process after 50 s, the control surface and engine loss pull are added. The pull conditions are set as shown in Table 2. The other simulation conditions are the same as in (6), and a comparative simulation analysis is performed.

[0051] Table 2. Setting of Pull-off Conditions

[0052] Figure 12-19Comparative results of dynamic inverse control based on cascaded observations are presented. It is evident that before rudder surface pull occurs, both the conventional dynamic inverse method and the cascaded observation dynamic inverse method can track altitude and lateral position commands well, with minimal difference in response. After 50 seconds, the conventional dynamic inverse method, affected by rudder effectiveness and engine loss, exhibits more pronounced fluctuations in the speed, angle of attack, pitch, and track tilt channels. Introducing the cascaded extended state observer allows the controller to estimate and compensate for equivalent disturbances, resulting in smoother responses in each state. Quantitatively, the maximum fluctuation amplitude of speed after engine loss is reduced by 46.99%; the maximum fluctuation amplitudes of angle of attack, pitch, and track tilt after 50 seconds are reduced by 50.30%, 57.27%, and 43.02%, respectively. Simultaneously, the peak responses of roll and sideslip angles are also suppressed, indicating that the cascaded observer not only improves longitudinal tracking accuracy but also enhances the adaptability of the lateral channels to control effectiveness losses. Furthermore, Figure 20-21 A comparison of torque accuracy and control surface response diagrams based on the weighted pseudo-inverse control allocation method are presented. It can be seen that the 12 control surfaces can achieve coordinated deflection, the three-channel decoupling torque accuracy is high, the control surfaces do not exceed the saturation boundary, and precise torque tracking is achieved. In summary, the nonlinear dynamic inverse decoupling method can improve control accuracy, the cascaded extended state observer can effectively compensate for the effects of control surface performance pull and model uncertainty, improving the robustness of the control system, and the weighted pseudo-inverse control allocation method can balance timeliness and accuracy. This verifies the effectiveness of the proposed dynamic inverse glide path control method for flying-wing UAVs based on cascaded observation.

Claims

1. A dynamic reverse glide control method for a flying-wing unmanned aerial vehicle based on cascaded observation, characterized in that, Includes the following steps: Step (1) Dynamic modeling of flying-wing UAV Step (2) Design of the guiding law for the trajectory loop To achieve trajectory glide control of a flying-wing UAV, it is first necessary to set the glide path and derive attitude loop control commands based on the trajectory loop guidance law. Step (3) Dynamic inverse control based on cascaded extended state observation Based on the attitude control command obtained in step (2), nonlinear dynamic inverse control of the attitude angle is first implemented, and then a cascaded extended state observer is introduced to achieve interference observation compensation. Step (4) Control allocation based on weighted pseudo-inverse The three-axis torque command obtained in step (3) is assigned to all control surfaces of the UAV to achieve control allocation.

2. The dynamic reverse glide control method for a flying-wing UAV based on cascaded observation as described in claim 1, characterized in that, Step (1) is as follows: The force and moment coefficients of the established flying-wing UAV dynamic model are shown below: (1) (2) In the formula: These are the body coordinate system x Directional force coefficient, y Directional force coefficient and z Directional force coefficient; These are the roll moment coefficient, pitch moment coefficient, and yaw moment coefficient, respectively. For the angle of attack, Sideslip angle, For speed, These are roll rate, pitch rate, and yaw rate, respectively. For control surface configuration, For thrust, The left inner leading edge flap and the right inner leading edge flap are the two flaps. The outer leading edge flaps are on the left and right sides. The left and right all-moving wingtips are respectively. The left and right elevons are the left and right elevons. The left and right pitch flaps are the pitch flaps. The rudder is for left-side drag and right-side drag; Based on the above expressions for force and moment coefficients, the expressions for force and moment of the flying-wing UAV are further obtained as follows: (3) In the formula: , , These are the body coordinate systems. x Directional force, y Directional force and z Directional force, , , These are respectively the rolling moment, pitching moment, and yaw moment. Atmospheric density; Wing area; For the average aerodynamic chord length, For wingspan, It is the thrust vector arm. The pitch thrust vector This is the yaw thrust vector; The dynamic model of the flying-wing UAV is shown below: (4) In the formula: superscript " " represents the first derivative; For quality; It is the acceleration due to gravity; They are respectively x,y,z Three-axis velocity components; These are the first derivatives of the roll angle, pitch angle, and yaw angle, respectively. These are roll angle, pitch angle, and yaw angle, respectively. for x,y,z Three-axis speed; It is a constant related to the moment of inertia.

3. The dynamic reverse glide control method for a flying-wing UAV based on cascaded observation as described in claim 1, characterized in that, Step (2) is as follows: Considering the control objectives during the descent phase, including longitudinal height tracking and lateral position convergence, the height error and lateral position error are defined as follows: (5) In the formula: This is the height command corresponding to the dynamic glide path; This refers to the actual height. This is for height error; For reference lateral position, This is the actual lateral position. This refers to lateral position error; Pitch and roll commands are generated from altitude and lateral position errors, respectively. (6) (7) In the formula: To balance the pitch angle; To control the gain; Indicates the instruction limiting function; This is the reference command for the roll angle; The reference command for the pitch angle; For the time of descent; This is a time parameter.

4. The dynamic reverse glide control method for a flying-wing UAV based on cascaded observation as described in claim 1, characterized in that, Step (3) is as follows: (3.1) Design of slow loop attitude angle control law Choose the slow loop variable as The slow loop corresponds to the attitude angle loop, resulting in the following set of nonlinear differential equations. (8) In the formula: For the lateral force of the wind axis; For engine thrust; The expression is as follows: (9) Attitude angular velocity vector is Considering the disturbances and model uncertainties during the descent, the uncertainties in the attitude angle loop are uniformly represented as equivalent disturbances. ,get: (10) in (11) (12) To improve the attitude angle loop's ability to compensate for external disturbances and model uncertainties, a cascaded extended state observer is introduced into the attitude angle loop; the virtual input of the attitude angle loop is defined. for (13) Then it is written as (14) Based on the structure of the cascaded extended state observer, the first-stage observer of the attitude angle loop is designed as follows: (15) The second-stage observer is designed as (16) In the formula: This refers to the attitude angle observation error; This is the estimated attitude angle value; Estimating disturbances for the first-level observer; Estimate the remaining perturbation for the second-stage observer; All are observer parameters; the total perturbation estimate for the attitude angle loop is... (17) in: Used to replace attitude angle measurements in error feedback Used to compensate for equivalent disturbances in the attitude angle loop; the observer parameters are calculated as follows: (18) In the formula: This represents the bandwidth of the attitude angle loop observer. It is a third-order identity matrix; rewrite formula (10) as follows: (19) In the formula: For the control parameters, we have the following expression. (20) when and Reversible, therefore the control law for the slow loop of the system is: (21)； (3.2) Design of fast-loop attitude angular velocity control law The fast loop corresponds to the torque equation of the UAV, and the attitude angular velocity loop. As a control input quantity As a control output, it includes all control surfaces; considering aerodynamic moment model errors, control surface performance deviations, and wind disturbances, the uncertainties in the angular velocity loop are uniformly represented as equivalent disturbances. The fast-loop angular rate equation is rearranged to obtain (22) make In the formula (23) (24) Define angular velocity loop virtual input for (25) Then it is written as (26) A cascaded extended state observer is introduced into the angular velocity loop. The first-stage observer is designed as follows: (27) The second-stage observer is designed as (28) In the formula: This refers to the angular velocity error. This is an estimate of the angular velocity; Estimating disturbances for the first-level observer; Estimate the remaining perturbation for the second-stage observer; All are observer parameters; the total perturbation estimate for the angular velocity loop is... (29) in: Used to replace angular velocity measurements in error feedback; Used to compensate for equivalent disturbances in the angular velocity loop; the parameters of the angular velocity loop observer are selected as follows: (30) In the formula: Let be the bandwidth of the angular velocity loop observer; according to formula (22), its calculation formula is as follows: (31) (32) In the formula: For the control parameters, we have the following expression. (33) The control law of the fast loop of the system is (34) (3.3) Dynamic compensation loop control design The automatic throttle module is designed to maintain speed, with the thrust vector angle set to 0, and a dynamic inverse method is used for the design. The differential formula for velocity is shown below: (35) In the formula: The wind axis resistance is used; by adjusting and deforming it, we obtain: (36) in (37) (38) In the formula: This represents the thrust corresponding to the current nominal throttle position. The partial derivative of thrust with respect to throttle command; This is the throttle command; Define the virtual input of the speed loop for (39) Written (40) A cascaded extended state observer is introduced into the velocity loop. The first-stage observer is designed as follows: (41) The second-stage observer is designed as (42) In the formula: For speed error; This is a speed estimate; Estimating perturbations for the first-level ESO; Estimate the remaining perturbations for the second-level ESO; All are observer parameters; the total perturbation estimate for the angular velocity loop is... (43) in: Used to replace speed measurements in error feedback; Used to compensate for equivalent disturbances in the velocity loop; the velocity loop observer parameters are selected as follows: (44) In the formula: For the velocity loop observer bandwidth; After introducing the cascaded observer, equation (35) is rewritten as follows: (45) In the formula: For the desired approach speed, For speed channel control gain; ultimately, the speed control law is: (46)。 5. The dynamic reverse glide control method for a flying-wing UAV based on cascaded observation as described in claim 1, characterized in that, Step (4) is as follows: Let the virtual control quantity of the system be (47) (48) The control effectiveness matrix is ​​denoted as Then the control allocation relationship is written as (49) Considering that different control surfaces have different travel limits, speed limits, and usage priorities, a weighted pseudo-inverse control allocation is adopted; a performance index function is defined. for (50) The constraints are (51) Then its optimal solution can be written as: (52) in (53) In the formula: Assign a matrix to the weighted pseudo-inverse control; The control surface weight matrix; through appropriate selection The allocation results of each control surface are optimized.

Images