A precise landing control method for vehicle-mounted UAV visual servoing considering field of view constraints

Abstract

The present invention discloses a method for precise landing control of a vehicle-mounted unmanned aerial vehicle (UAV) using visual servoing, taking into account field-of-view constraints. The method comprises the following steps: using virtual plane feature points to construct image moment eigenvalues, establishing a decoupled system dynamics model that integrates UAV dynamics and image dynamics; obtaining the current system state based on the image captured by the camera and the UAV's motion state; setting outer and inner control loops based on a model predictive control method, wherein an optimization target is set in the outer control loop, and attitude constraints are set in the inner control loop. Based on the current system state, the system state at subsequent moments is predicted, and the optimal control at the current moment is obtained as system input; and flight control of the UAV is performed based on the system input. The method disclosed by the present invention ensures that field-of-view visibility is met during the autonomous landing of the UAV.

Description

Technical Field

[0001] The present invention relates to a vehicle-mounted unmanned aerial vehicle (UAV) visual servo precise landing control method considering field of view constraints, and belongs to the technical field of aircraft control. Background Art

[0002] The visual servoing method can still obtain navigation information in a GPS-denied environment, and the visual positioning method is a non-contact measurement with good reliability. Therefore, it is often used in autonomous flight control of UAVs.

[0003] Most existing visual servo control methods do not consider the constraints of visibility conditions or the impact of the drone's posture on the image, which often causes the target to appear outside the camera's field of view during actual flight, making it impossible to obtain position information relative to the target, resulting in control failure.

[0004] Although there are visual servo control methods that consider visibility constraints in the existing technology, for example, in the literature [K. Zhang, Y. Shi and H. Sheng,"Robust Nonlinear Model Predictive Control Based Visual Servoing of Quadrotor UAVs," in IEEE / ASME Transactions on Mechatronics, vol. 26, no. 2, pp. 700-708, April 2021, doi: 10.1109 / TMECH.2021.3053267], a nonlinear model predictive control method (NMPC) is used to achieve robust control of autonomous landing of UAVs, but its visibility constraint is analyzed based on the virtual image plane; in the literature [H. Sheng, E. Shi and K. Zhang,"Image-Based Visual Servoing of a Quadrotor with Improved Visibility Using Model Predictive Control," 2019 IEEE 28th International Symposium on Industrial Electronics(ISIE),Vancouver,BC,Canada,2019,pp.551-556,doi:10.1109 / ISIE.2019.8781212] A visual servo controller based on MPC with field of view constraints is designed for UAV landing control. However, directly introducing attitude angle constraints may result in the failure to meet the field of view constraints. It can be seen that the existing visual servoing method still cannot guarantee control stability.

[0005] Therefore, it is necessary to conduct more in-depth research on existing visual servoing methods to solve the above problems. Summary of the Invention

[0006] In order to overcome the above problems, the inventors conducted in-depth research and proposed a vehicle-mounted UAV visual servoing precise landing control method considering field of view constraints, comprising the following steps: S1, using virtual plane feature points to construct image moment eigenvalues, and establishing a decoupled system dynamics model that integrates UAV dynamics and image dynamics;

[0007] S2. Obtain the current system status based on the image captured by the camera and the motion status of the drone;

[0008] S3. Setting an outer-loop control and an inner-loop control based on a model predictive control method, wherein an optimization target is set in the outer-loop control and a posture constraint is set in the inner-loop control. Based on the current system state, the system state at subsequent moments is predicted to obtain the optimal control at the current moment as the system input;

[0009] S4. Based on the system input, the UAV is controlled in flight.

[0010] In a preferred embodiment, in S1, the virtual plane is a plane established between the drone and the horizontal plane and parallel to the horizontal plane.

[0011] In a preferred embodiment, the feature point is an intersection point of the edges of any square on the virtual plane.

[0012] In a preferred embodiment, the image moment eigenvalue s is set to

[0013] s=[s1,s2,s3,s4]

[0014]

[0015] Among them, s1, s2, s3, and s4 are intermediate variables, [u vc ,n vc ] represents the coordinates of the center of gravity of the feature point, f is the focal length of the camera, a represents the acceleration of the drone, a d Desired altitude for the drone The corresponding a value, u r1 Indicates the horizontal coordinate of the first point corresponding to the feature point in the real image, n r1 Indicates the vertical coordinate of the first point corresponding to the feature point in the real image, u r2 Indicates the horizontal coordinate of the second point corresponding to the feature point in the real image, n r2 Indicates the ordinate of the second point corresponding to the feature point in the real image, u rcIndicates the horizontal coordinates of the center of gravity of the four points corresponding to the feature points in the real image, n rc Indicates the horizontal coordinates of the centroid of the four points corresponding to the feature points in the real image.

[0016] In a preferred embodiment, the image dynamics is represented as:

[0017]

[0018] in, represents the velocity of the UAV in the virtual camera coordinate system, and ψ represents the yaw angle of the UAV.

[0019] In a preferred embodiment, the UAV dynamics are expressed as:

[0020]

[0021] Among them, p represents the position of the UAV in the ground system, v represents the speed of the UAV in the ground system, represents the rotation matrix between the drone system and the inertial system, T represents the total lift generated by the blades, and E3 = [0,0,1] T represents the direction of the lift force T in the drone system, m represents the mass of the drone, g represents the acceleration due to gravity, [·] × represents the operator that converts a vector into a skew-symmetric matrix, Ω represents the angular velocity of the drone, I represents the moment of inertia of the drone, and τ represents the input torque.

[0022] In a preferred embodiment, the decoupled system dynamics model is expressed as:

[0023]

[0024] Where x represents the system state, is the first-order derivative of the system state x, and u represents the system input.

[0025] In a preferred embodiment, the outer loop control is configured as follows:

[0026]

[0027] u(k+i|k)∈u set

[0028] in, is the objective function of predictive control, To optimize the indicators, represents the process cost function, represents the terminal cost function, x drepresents the expected system state, x(k+i|k) represents the prediction of the system state at step k+i based on the system input u(k|k) at step k; It represents the error between the prediction made at step k for step k+i and the expected value, represents the first-order derivative of the prediction made by step k for step k+i, u(k+i|k) represents the system input of step k+i predicted by step k, i∈{0,···,N}, N is the length of the prediction time step interval, u set Indicates the value range of u.

[0029] In a preferred embodiment, the inner loop control is configured as follows:

[0030]

[0031] e Ω =Ω-R T R d Ω d

[0032] τ=-k R e R -k Ω e Ω +Ω^JΩ

[0033] θ d =θ-e R (2)

[0034] φ d =φ-e R (1)

[0035] b=[cos(θ),sin(θ),0]′

[0036] Among them, R d is the desired rotation matrix, r d1 , r d2 , r d3 is the component of the desired rotation matrix, b is the intermediate parameter, Ω d is the intermediate matrix of angular velocity, e R is the deviation between the expected attitude angle and the actual attitude angle, e Ω represents the deviation between the expected angular velocity and the actual angular velocity, τ represents the torque of the drone, and k R 、k Ω is a settable parameter, J is the moment of inertia, the superscript T indicates transpose, and the superscripts ^ and ∨ indicate matrix transformation symbols. ^ indicates transforming a 3*1 matrix into a 3*3 matrix, and ∨ indicates restoring a 3*3 matrix into a 3*1 matrix.

[0037] In a preferred embodiment, the maximum allowable pitch angle θ is set to max, Maximum allowable roll angle φ max Perform posture constraints:

[0038] When θ max <θ d When, e R (2) = -(θ max -θ), otherwise e R (2) is the deviation between the desired pitch angle and the actual pitch angle;

[0039] φ max <φ d When, e R (1)=-(φ max -φ), otherwise e R (1) is the deviation between the desired roll angle and the actual roll angle.

[0040] The beneficial effects of the present invention include:

[0041] (1) Ensure the visibility of the UAV during autonomous landing;

[0042] (2) High control stability and high landing accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of a method for controlling a vehicle-mounted UAV's visual servoing precise landing with consideration of field of view constraints according to a preferred embodiment of the present invention is shown;

[0044] Figure 2 The horizontal position change curve of the UAV in the static target landing simulation of Example 1 is shown;

[0045] Figure 3 The vertical position change curve of the UAV in the static target landing simulation of Example 1 is shown;

[0046] Figure 4 The three-dimensional tracking trajectory of the UAV landing in the static target landing simulation of Example 1 is shown;

[0047] Figure 5 The figure shows the change of the X-axis coordinates of the image feature points in the static target landing simulation of Example 1;

[0048] Figure 6 The figure shows the change in pitch angle of the UAV in the static target landing simulation of Example 1;

[0049] Figure 7 The horizontal position change curve of the UAV in the dynamic target landing simulation of Example 1 is shown;

[0050] Figure 8 The vertical position change curve of the UAV in the dynamic target landing simulation of Example 1 is shown;

[0051] Figure 9 The three-dimensional tracking trajectory of the UAV landing in the dynamic target landing simulation of Example 1 is shown;

[0052] Figure 10 The figure shows the change of the X-axis coordinates of the image feature points in the dynamic target landing simulation of Example 1;

[0053] Figure 11 The figure shows the change in pitch angle of the drone in the dynamic target landing simulation of Example 1;

[0054] Figure 12 The figure shows the change of the X-axis coordinates of the image feature points in the static target landing simulation of Example 2;

[0055] Figure 13 The figure shows the change in the pitch angle of the drone in the static target landing simulation of Example 2;

[0056] Figure 14 The figure shows the change of the X-axis coordinates of the image feature points in the dynamic target landing simulation of Example 2;

[0057] Figure 15 The figure shows the change of the pitch angle of the UAV in the dynamic target landing simulation of Example 2. DETAILED DESCRIPTION

[0058] The present invention will be described in further detail below with reference to the accompanying drawings and examples, through which the features and advantages of the present invention will become more clearly understood.

[0059] The word "exemplary" is used exclusively herein to mean "serving as an example, example, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0060] According to the present invention, a vehicle-mounted UAV visual servoing precise landing control method considering field of view constraints is provided, which realizes the autonomous landing of the UAV for stationary and moving targets. Figure 1 As shown, including:

[0061] S1. Using virtual plane feature points to construct image moment eigenvalues, a decoupled system dynamics model integrating UAV dynamics and image dynamics is established.

[0062] S2. Obtain the current system status based on the image captured by the camera and the motion status of the drone;

[0063] S3. Setting an outer-loop control and an inner-loop control based on a model predictive control method, wherein an optimization target is set in the outer-loop control and a posture constraint is set in the inner-loop control. Based on the current system state, the system state at subsequent moments is predicted to obtain the optimal control at the current moment as the system input;

[0064] S4. Based on the system input, the UAV is controlled in flight.

[0065] In S1, the virtual plane is a plane parallel to the horizontal plane established between the drone and the horizontal plane. In the present invention, by setting this virtual plane, the influence of the drone's posture can be eliminated in the subsequent solution process, reducing variables and thus reducing the amount of calculation.

[0066] In the present invention, the feature point may be an intersection point of edges of any regular figure on a virtual plane.

[0067] In a preferred embodiment, the feature points are the intersection points of the side lines of any square on the virtual plane. The four intersection points of the side lines of the square are easier to locate and perform subsequent calculations than other points.

[0068] The image moment eigenvalue s is set to

[0069] s=[s1,s2,s3,s4]

[0070]

[0071] Among them, s1, s2, s3, and s4 are intermediate variables, [u vc ,n vc ] represents the coordinates of the center of gravity of the feature point, f is the focal length of the camera, a represents the acceleration of the drone, a d Desired altitude for the drone The corresponding a value, u r1 Indicates the horizontal coordinate of the first point corresponding to the feature point in the real image, n r1 Indicates the vertical coordinate of the first point corresponding to the feature point in the real image, u r2 Indicates the horizontal coordinate of the second point corresponding to the feature point in the real image, n r2 Indicates the ordinate of the second point corresponding to the feature point in the real image, u rc Indicates the horizontal coordinates of the center of gravity of the four points corresponding to the feature points in the real image, n rc Represents the horizontal coordinates of the centroid of the four points corresponding to the feature points in the real image.

[0072] Furthermore, a=μ 20 +μ 02

[0073] Among them, μ 20 、μ 02Obtained by the following formula:

[0074]

[0075] Among them, l represents different feature points, represents the horizontal coordinate of the lth feature point, Represents the average value of the horizontal coordinates of the four feature points, The ordinate of the lth feature point, Represents the average value of the vertical coordinates of the four feature points, i,j∈[0,2].

[0076] The image dynamics is expressed as:

[0077]

[0078] in, represents the velocity of the UAV in the virtual camera coordinate system, and ψ represents the yaw angle of the UAV.

[0079] The virtual camera coordinate system is retained with the machine system F b In F n The coordinate system of the UAV is the same as the translational motion in the original image, but the rotational motion of the UAV is not preserved.

[0080] Furthermore, combining coordinate transformation and drone dynamics, we have:

[0081]

[0082] The UAV dynamics is expressed as:

[0083]

[0084] Among them, p represents the position of the UAV in the ground system, v represents the speed of the UAV in the ground system, represents the rotation matrix between the drone system and the inertial system, T represents the total lift generated by the blades, and E3 = [0,0,1] T represents the direction of the lift force T in the drone system, m represents the mass of the drone, g represents the acceleration due to gravity, [·] × represents the operator that converts a vector into a skew-symmetric matrix, Ω represents the angular velocity of the drone, I represents the moment of inertia of the drone, and τ represents the input torque.

[0085] Based on the image moment eigenvalues, drone dynamics and image dynamics, a decoupled system dynamics model can be obtained. The system dynamics model is based on the image moment eigenvalues ​​s and the velocity v of the drone in the virtual camera coordinate system. v Get the system state, with acceleration and yaw rate as system input.

[0086] The decoupled system dynamics model is expressed as:

[0087]

[0088] Where x represents the system state, is the first-order derivative of the system state x, and u represents the system input.

[0089] Furthermore, the system input is expressed as Among them, a x ,a y ,a z are the components of the drone's acceleration in the x, y, and z directions.

[0090] Furthermore, the lift F of the drone can be obtained according to the acceleration k , expressed as:

[0091]

[0092] Where m is the mass of the drone.

[0093] Compared with the traditional dynamics model, the decoupled system dynamics model in the present invention can directly analyze the system dynamics characteristics based on the virtual plane without considering the influence of the drone's posture on the dynamics.

[0094] In S2, the image moment eigenvalue s is obtained according to the image taken by the camera, and the speed of the drone in the virtual camera coordinate system is obtained according to the speed of the drone.

[0095] In S3, model predictive control (MCP) is a commonly used predictive control method. In the present invention, the model predictive control includes outer loop control and inner loop control.

[0096] Wherein, the outer loop control is set as:

[0097]

[0098]

[0099] u(k+i|k)∈u set

[0100] in, is the objective function of predictive control, To optimize the indicators, represents the process cost function, represents the terminal cost function, x d represents the expected system state, x(k+i|k) represents the prediction of the system state at step k+i based on the system input u(k|k) at step k; It represents the error between the prediction made at step k for step k+i and the expected value, represents the first-order derivative of the prediction made by step k for step k+i, u(k+i|k) represents the system input of step k+i predicted by step k, i∈{0,···,N}, N is the length of the prediction time step interval, u set It represents the value range of u, and the specific range can be freely set by those skilled in the art according to actual needs and is not limited in the present invention.

[0101] Preferably, the process cost function and the terminal cost function are set to:

[0102]

[0103] Among them, U k To control the queue, is the error queue, expressed as:

[0104] U k ={u k|k T ,u k+1|k T ,……,u k+N|k T} T

[0105]

[0106] The superscript T indicates transpose, and Q, R, and F are configurable matrix parameters.

[0107] In a preferred embodiment, considering the differences in the detection effects of cameras at target points at different heights in actual scenes, the landing process is divided into two stages according to the landing height:

[0108] When the drone's altitude exceeds the preset altitude When , the values ​​of Q, R, and F are set to Q1, R1, and F1 respectively;

[0109] When the drone's altitude is lower than the preset altitude When , the values ​​of Q, R, and F are set to Q2, R2, and F2 respectively.

[0110] In the present invention, there is no limitation on the specific values ​​of Q1, R1, F1, Q2, R2, and F2, and technicians in this field can freely set them according to actual needs.

[0111] According to the present invention, the inner loop control is set as follows:

[0112]

[0113] e Ω =Ω-R TR d Ω d

[0114] τ=-k R e R -k Ω e Ω +Ω^JΩ

[0115] θ d =θ-e R (2)

[0116] φ d =φ-e R (1)

[0117] b=[cos(θ),sin(θ),0]′

[0118] Among them, R d is the desired rotation matrix, r d1 , r d2 , r d3 is the component of the desired rotation matrix, b is the intermediate parameter, Ω d is the intermediate matrix of angular velocity, e R is the deviation between the expected attitude angle and the actual attitude angle, e Ω represents the deviation between the expected angular velocity and the actual angular velocity, τ represents the torque of the drone, and k R 、k Ω is a settable parameter, J is the moment of inertia, the superscript T indicates transpose, and the superscripts ^ and ∨ indicate matrix transformation symbols. ^ indicates transforming a 3*1 vector into a 3*3 matrix, and ∨ indicates restoring a 3*3 matrix into a 3*1 vector.

[0119] Examples of superscript ^ operation:

[0120] For example, a=[a1,a2,a3] T , a ∧ Then A ∨ Then it is a.

[0121] According to the present invention, the maximum allowable pitch angle θ is set max , Maximum allowable roll angle φ max Perform posture constraints:

[0122] When θ max <θ d When, e R (2) = -(θ max -θ), otherwise e R (2) is the deviation between the desired pitch angle and the actual pitch angle;

[0123] φ max <φ d When, eR (1)=-(φ max -φ), otherwise e R (1) is the deviation between the desired roll angle and the actual roll angle.

[0124] Preferably, the maximum allowable pitch angle θ max Obtained through:

[0125]

[0126] in, Indicates the distance between the image feature point and the image center, cd is the length corresponding to when the feature point is 1, β is the intermediate parameter, x k+1 (1) represents the first element of the predicted system state at step k+1, x k+1 (3) represents the third element of the system state quantity in the predicted k+1th step.

[0127] Preferably, the maximum allowable roll angle φ max Obtained through:

[0128]

[0129] in, is the intermediate parameter, x k+1 (2) represents the second element of the system state quantity in the predicted k+1th step.

[0130] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved. This is not limited herein.

[0131] Example

[0132] Example 1

[0133] Conduct UAV visual servo landing control simulation experiments, including:

[0134] S1. Using virtual plane feature points to construct image moment eigenvalues, a decoupled system dynamics model integrating UAV dynamics and image dynamics is established.

[0135] S2. Obtain the current system status based on the image captured by the camera and the motion status of the drone;

[0136] S3. Setting an outer-loop control and an inner-loop control based on a model predictive control method, wherein an optimization target is set in the outer-loop control and a posture constraint is set in the inner-loop control. Based on the current system state, the system state at subsequent moments is predicted to obtain the optimal control at the current moment as the system input;

[0137] S4. Based on the system input, the UAV is controlled in flight.

[0138] In S1, the virtual plane is a plane parallel to the horizontal plane established between the drone and the horizontal plane, and the image moment eigenvalue s is set to

[0139] s=[s1,s2,s3,s4]

[0140]

[0141] The image dynamics is expressed as:

[0142]

[0143] The UAV dynamics is expressed as:

[0144]

[0145] The decoupled system dynamics model is expressed as:

[0146]

[0147] The system input is expressed as

[0148] The lift force F of the drone k Expressed as

[0149] In S3, the outer loop control is set to:

[0150]

[0151] u(k+i|k)∈u set

[0152] The process cost function and terminal cost function are set as:

[0153]

[0154] The landing process is divided into two stages according to the landing altitude:

[0155] When the drone's altitude exceeds the preset altitude When , the values ​​of Q, R, and F are set to Q1, R1, and F1 respectively;

[0156] When the drone's altitude is lower than the preset altitude When , the values ​​of Q, R, and F are set to Q2, R2, and F2 respectively.

[0157] The inner loop control is set as:

[0158]

[0159]

[0160] e Ω =Ω-R T R d Ω d

[0161] τ=-k R e R -k Ω e Ω +Ω^JΩ

[0162] θ d =θ-e R (2)

[0163] φ d =φ-e R (1)

[0164] b=[cos(θ),sin(θ),0]′

[0165] Set the maximum allowable pitch angle θ max , Maximum allowable roll angle φ max Perform posture constraints:

[0166] When θ max <θ d When, e R (2) = -(θ max -θ), otherwise e R (2) is the deviation between the desired pitch angle and the actual pitch angle;

[0167] φ max <φ d When, e R (1)=-(φ max -φ), otherwise e R (1) is the deviation between the desired roll angle and the actual roll angle.

[0168] Maximum allowable pitch angle θ max Obtained through:

[0169]

[0170] Maximum allowable roll angle φ max Obtained through:

[0171]

[0172]

[0173] In the simulation experiment, the coordinates of the four feature points in the virtual plane in the first stage of landing are P 11 =(6,1,0)m,P 12 =(4,1,0)m,P 13 =(4,-1,0)m,

[0174] P 14 =(6,-1,0)m, the preset height is set to Set the expected moment feature a′ d =14.22; In the second stage, the coordinates of the four feature points in the virtual plane are selected as P 21 =(5.1,0.1,0)m,P 22 =(4.9,0.1,0)m,P 23 =(4.9,-0.1,0)m,

[0175] P 24 =(5.1,-0.1,0)m, the preset height is set to The corresponding expected moment feature a d =1.5802; prediction interval N=3, sampling time dt=0.001s. The mass m of the drone is 1.8kg, and the moment of inertia is

[0176] J=diag(0.01562,0.01456,0.02492)kgm 2 , g=9.81m / s 2 , camera parameters f = 0.02m, cd = 0.015m. The weight matrices for the first stage are Q1 = diag(60,60,60,200,1,0.001,0.001,0.001), R1 = diag(0.005,0.005,0.005,0.005), and F1 = diag(5,5,200,1,5,5,5). The weight matrices for the second stage are Q2 = diag(18,18,0.1,18,0.001,0.001,0.001), R2 = diag(0.005,0.005,0.005,0.005), and F2 = diag(3,3,10,3,3,3,3). Set the desired state vector to x d =[0,0,1,0,0,0,0].

[0177] During the simulation process, static target landing simulation and dynamic target landing simulation are set. The static target landing simulation results are as follows: Figure 2-Figure 6 As shown,

[0178] Figure 2 The horizontal position change curve of the UAV is shown. Figure 3 The vertical position change curve of the UAV is shown. Figure 4 The three-dimensional tracking trajectory of the UAV landing is shown. Figure 2-4 It can be seen that the horizontal and vertical positions of the drone have converged to the desired positions;

[0179] Figure 5 The X-axis coordinate changes of the image feature points are shown. It can be seen from the figure that the outermost feature points do not exceed the field of view, so the field of view visibility constraint is met;

[0180] Figure 6 The figure shows the pitch angle change of the drone. It can be seen from the figure that the actual pitch angle is strictly limited to the maximum pitch angle range, so the landing meets the attitude angle constraint;

[0181] Dynamic target landing simulation results are as follows Figure 7-11 As shown,

[0182] Figure 7 The horizontal position change curve of the UAV is shown. Figure 8 The vertical position change curve of the UAV is shown. Figure 9 The three-dimensional tracking trajectory of the UAV landing is shown. Figure 7-9 It can be seen that the horizontal and vertical positions of the drone have converged to the desired positions;

[0183] Figure 10 The X-axis coordinate changes of the image feature points are shown. It can be seen from the figure that the outermost feature points do not exceed the field of view, so the field of view visibility constraint is met;

[0184] Figure 11 The figure shows the pitch angle change of the drone. It can be seen from the figure that the actual pitch angle is strictly limited to the range of the maximum pitch angle, so the landing meets the attitude angle constraint.

[0185] Example 2

[0186] The same experiment as in Example 1 was conducted, except that the maximum allowable pitch angle θ was not set. max , Maximum allowable roll angle φ max Perform posture constraints.

[0187] In the static target landing simulation results, the X-axis coordinates of the image feature points change as follows: Figure 12 As shown in the figure, it can be seen that the position of the feature point is outside the field of view;

[0188] The pitch angle of the drone changes as follows Figure 13 As shown in the figure, it can be seen that the actual pitch angle is not within the range of the maximum allowable pitch angle, and the actual pitch angle and the maximum allowable pitch angle intersect in the initial stage of landing;

[0189] In the dynamic target landing simulation results, the X-axis coordinates of the image feature points change as follows: Figure 14 As shown in the figure, it can be seen that the position of the feature point is outside the field of view, and the position of the feature point in the image reaches the boundary of the field of view at an earlier time, about 0.45s;

[0190] The pitch angle of the drone changes as follows Figure 15 As shown in the figure, it can be seen that the actual pitch angle curve and the maximum allowable pitch angle curve intersect in the initial stage of landing, which means that the field of view visibility condition is not met.

[0191] Comparative Example 1 Figure 5 、 Figure 6 、 Figure 10 、 Figure 11 With Example 2 Figure 12 、 Figure 13 、 Figure 14 、 Figure 15 It can be clearly seen that in Example 1, by setting the maximum allowable pitch angle θ max , Maximum allowable roll angle φ max Performing attitude constraints can effectively ensure the visibility of the field of view during the autonomous landing of the UAV, and improve the control stability and landing accuracy of the UAV.

[0192] The present invention has been described above with reference to preferred embodiments, but these embodiments are merely exemplary and serve only as illustrations. On this basis, various replacements and improvements can be made to the present invention, all of which fall within the scope of protection of the present invention.

Claims

1. A method for precise landing control of a vehicle-mounted UAV visual servo system considering field of view constraints, characterized in that: The following steps are involved: S1. Using virtual plane feature points to construct image moment eigenvalues, a decoupled system dynamics model integrating UAV dynamics and image dynamics is established. S2. Obtain the current system status based on the image captured by the camera and the motion status of the drone; S3. Setting an outer-loop control and an inner-loop control based on a model predictive control method, wherein an optimization target is set in the outer-loop control and a posture constraint is set in the inner-loop control. Based on the current system state, the system state at subsequent moments is predicted to obtain the optimal control at the current moment as the system input; S4, based on the system input, control the flight of the UAV; The outer loop control is set as: u(k+i|k)∈u set in, is the objective function of predictive control, To optimize the indicators, represents the process cost function, represents the terminal cost function, x d represents the expected system state, x(k+i|k) represents the prediction of the system state at step k+i based on the system input u(k|k) at step k; It represents the error between the prediction made at step k for step k+i and the expected value, represents the first-order derivative of the prediction made by step k for step k+i, u(k+i|k) represents the system input of step k+i predicted by step k, i∈{0,···,N}, N is the length of the prediction time step interval, u set Indicates the value range of input u; The inner loop control is set as: e Ω =Ω-R T R d Oh d τ=-k R a R -k Ω a Ω +Ω^JΩ i d =θ-e R (2) f d =φ-e R (1) b=[cos(θ),sin(θ),0]′ Among them, R d is the desired rotation matrix, r d1 ,r d2 ,r d3 is the component of the desired rotation matrix, b is the intermediate parameter, Ω d is the intermediate matrix of angular velocity, e R is the deviation between the expected attitude angle and the actual attitude angle, e Ω represents the deviation between the expected angular velocity and the actual angular velocity, τ represents the torque of the drone, and k R 、k Ω is a settable parameter, J is the moment of inertia, the superscript T represents transpose, the superscripts ^ and ∨ represent matrix transformations, ^ represents transforming a 3*1 vector into a 3*3 matrix, and ∨ represents restoring a 3*3 matrix into a 3*1 vector. k represents the lift of the UAV, and Ω represents the angular velocity of the UAV.

2. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 1 is characterized in that: In S1, the virtual plane is a plane established between the drone and the horizontal plane and parallel to the horizontal plane.

3. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 1 is characterized in that: The feature point is the intersection point of the side lines of any square on the virtual plane.

4. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 1 is characterized in that: The image moment eigenvalue s is set to s=[S1, S2, S3, S4] Among them, s1, s2, s3, and s4 are intermediate variables, [u vc ,n vc ] represents the coordinates of the center of gravity of the feature point, f is the focal length of the camera, a represents the acceleration of the drone, a d Desired altitude for the drone The corresponding a value, u r1 Indicates the horizontal coordinate of the first point corresponding to the feature point in the real image, n r1 Indicates the vertical coordinate of the first point corresponding to the feature point in the real image, u r2 Indicates the horizontal coordinate of the second point corresponding to the feature point in the real image, n r2 Indicates the ordinate of the second point corresponding to the feature point in the real image, u rc Indicates the horizontal coordinates of the center of gravity of the four feature points in the real image, n rc Represents the centroid ordinates of the four feature points corresponding to the real image.

5. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 4 is characterized in that: The image dynamics is expressed as: in, represents the velocity of the UAV in the virtual camera coordinate system, and ψ represents the yaw angle of the UAV.

6. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 1 is characterized in that: The UAV dynamics is expressed as: Among them, p represents the position of the UAV in the ground system, v represents the speed of the UAV in the ground system, represents the rotation matrix between the drone system and the inertial system, T represents the total lift generated by the blades, and E3 = [0,0,1] T represents the direction of the lift force T in the drone system, m represents the mass of the drone, g represents the acceleration due to gravity, [·] × represents the operator that converts a vector into a skew-symmetric matrix, I represents the moment of inertia of the drone, and τ represents the input torque.

7. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 5 is characterized in that: The decoupled system dynamics model is expressed as: Where x represents the system state, is the first-order derivative of the system state x, and u represents the system input.

8. The method for precise landing control of a vehicle-mounted UAV with visual servoing considering field of view constraints according to claim 1 is characterized in that: Set the maximum allowable pitch angle θ max , Maximum allowable roll angle φ max Perform posture constraints: When θ max <θ d When, e R (2) = -(θ max -θ), otherwise e R (2) is the deviation between the desired pitch angle and the actual pitch angle; φ max <φ d When, e R (1)=-(φ max -φ), otherwise e R (1) is the deviation between the desired roll angle and the actual roll angle.

Images