A UAV fixed-point spraying operation method based on KLMS algorithm

Abstract

The present invention discloses a method for fixed-point spraying operation of an unmanned aerial vehicle (UAV) based on the KLMS algorithm, comprising: the UAV carrying a laser transmitter is lifted into the air and performs laser marking to confirm the coordinates of the spray landing point; when the coordinates of the spray landing point are confirmed, the initial operating position of the UAV is calculated in combination with preset parameters including the horizontal operating distance and the spraying speed; the UAV reaches the initial operating position and starts the spraying operation after being permitted by ground monitoring personnel; during the operation, the KLMS-based landing point self-stabilization algorithm is executed to adjust the attitude and orientation of the UAV to ensure the stability of the spray landing point. The present invention focuses on the stability of the spray landing point to meet the needs of actual working scenarios. Taking into account the disturbances such as the recoil force generated by the UAV fixed-point spraying operation, in order to achieve a stable landing point, the KLMS-based landing point self-stabilization algorithm is used to adjust the attitude and orientation of the UAV, which greatly improves the positioning accuracy of the spraying operation and the efficiency of the spraying operation.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) operations, and in particular to a UAV fixed-point spraying operation method based on a KLMS algorithm. Background Art

[0002] In recent years, the research and application of drone patrol technology has experienced explosive growth, with applications expanding from traditional patrols to high-altitude operations such as live flame spraying, live water washing, and compressed air spraying. It is generally believed that the disturbances received by drones during flight are random and have significant nonlinear characteristics. Existing control strategies stabilize drones at specific positions and attitudes, which do not meet the needs of actual work scenarios. Spraying operations suffer from poor positioning accuracy and efficiency, and fail to achieve the expected operational results. Furthermore, stabilization control methods based on classic PID are inefficient and ineffective. Although backstepping, a nonlinear control method, has been used to design attitude controllers, achieving better control results, deviations can occur in some cases, even leading to overcorrection and exacerbated vibrations. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a UAV fixed-point spraying operation method based on the KLMS algorithm.

[0004] In order to solve the above technical problems, the present invention provides a UAV fixed-point spraying operation method based on the KLMS algorithm, comprising:

[0005] The drone carries a laser transmitter and takes off into the air to perform laser marking to confirm the coordinates of the jet landing point;

[0006] After confirming the coordinates of the spraying point, the drone's operating position is calculated based on the preset parameters including the horizontal operating distance and spraying speed.

[0007] The UAV arrives at the operation location and begins spraying after being approved by the ground supervisor;

[0008] During the operation, it is continuously confirmed whether the drone is within the electronic fence. If so, the KLMS-based landing point self-stabilization algorithm is executed to adjust the drone's attitude and orientation to ensure the stability of the spraying landing point. If not, the drone is controlled to return to the initial operation point and continue the spraying operation.

[0009] Furthermore, the electronic fence is created with the calculated working position as the center.

[0010] Furthermore, a water carrying mode in which the drone and the water tank are separated is adopted. Specifically, the water tank is placed on the ground, one end of the water pipe is connected to the water tank, and the other end is towed by the drone.

[0011] Furthermore, a spray mode with a fixed spray angle is adopted.

[0012] Furthermore, the KLMS-based landing point self-stabilization algorithm is executed to adjust the attitude and orientation of the UAV to ensure the stability of the jet landing point, specifically including:

[0013] Obtain the position and attitude angle information of the UAV under the influence of disturbance, and calculate the actual injection point d(n) based on the ejection;

[0014] Calculate the error e(n) between the actual injection point d(n) and the theoretical injection point x(n);

[0015] Update the weight coefficient w(n) through the error e(n) and the convergence coefficient μ;

[0016] The updated weight coefficient w(n) is applied to the UAV dynamics model to change the UAV's power, thereby adjusting the UAV's attitude and orientation and predicting the UAV's position at the next moment.

[0017] Furthermore, the KLMS algorithm introduces the Gaussian kernel function based on the LMS algorithm to map the high-dimensional original input data {x(1), x(2),…, x(N)} into a high-dimensional space. Realize the transformation of nonlinear perturbation problems into linear problems;

[0018] Assume there is a mapping X→H maps x(i) to The corresponding output is {y(1), y(2), …, y(N)}. The iterative process of the KLMS algorithm is:

[0019] w(0)=0

[0020]

[0021]

[0022] Where e(n) is the error, d(n) is the actual injection point, μ is the convergence coefficient, and w(n) is the weight coefficient vector when the iteration is performed to the nth time, which is recorded as:

[0023]

[0024] Let the kernel function be:

[0025]

[0026] Then there are estimated values:

[0027]

[0028] The output of the KMLS algorithm is independent of the weight coefficient. By establishing a kernel function, multiplying it with the corresponding error and then linearly combining them, the complete output of the filter is obtained, thereby realizing the dimensionality reduction processing of nonlinear filtering.

[0029] Furthermore, if the drone's position crosses the electronic fence at the next moment, the drone will be controlled to return to the electronic fence first; if the drone's position does not cross the electronic fence at the next moment, the drone will continue to predict the injection landing point and complete the next cycle iteration.

[0030] Furthermore, the electronic fence is created with the initial operating position of the drone as the center.

[0031] Furthermore, the calculation of the injection landing point includes:

[0032] A right-handed coordinate system is established with the laser point as the origin P0(0,0,0), and the Y-axis direction is along the overhead line. At a certain moment, the drone is located at P(x0,y0,z0), and its attitude angles are The jet velocity is v. Assuming that there is no air resistance during the jetting process, the motion of the jetted liquid in the air at a certain moment is oblique projectile motion.

[0033] According to the laws of motion, the displacement of water when it reaches the YOZ plane is calculated, thereby predicting the jet landing point at a certain moment:

[0034]

[0035] Among them, x0 is the X-axis coordinate, is the roll angle around the X axis, θ is the pitch angle around the Y axis, ψ is the yaw angle around the Z axis, and g is the gravity constant.

[0036] Furthermore, the UAV dynamics model is as follows;

[0037]

[0038] in, is the second-order derivative of any variable x in the x-axis direction, is the second-order derivative of any variable y in the y-axis direction, is the second-order derivative of any variable z in the z-axis direction, is the first-order derivative of the yaw angle, is the first derivative of the pitch angle, is the first derivative of the roll angle, They are the control quantities of the three directions of the angular velocity of the quadrotor drone, U1 is the virtual control quantity of lift, U2, U3, and U4 are the virtual control quantities of roll angle, pitch angle, and yaw angle respectively, J1, J2, and J3 are the control quantities of the three directions of moment of inertia respectively, l is the distance between the center of mass of the drone and the rotor motor, and m is the mass of the drone.

[0039] The beneficial effects of implementing the present invention are:

[0040] 1. Different from the control ideas of existing drones, this invention focuses more on the stability of the jet landing point to meet the needs of actual working scenarios.

[0041] 2. Taking into account the disturbances such as the recoil force generated by the UAV's fixed-point spraying operation, in order to achieve a stable landing point, a landing point self-stabilization algorithm based on KLMS is used to adjust the UAV's attitude and orientation, which greatly improves the positioning accuracy and efficiency of the spraying operation.

[0042] 3. Using a separate water-carrying mode with the drone and water tank reduces the drone's load, ensuring optimal operation time. Furthermore, the interference from the water pipe on the drone can be roughly assumed to be constant, facilitating algorithmic offset. However, when operating with a water tank, the drone's center of gravity shifts as the water level decreases, causing the drone's mathematical model to deviate from actual conditions and weakening its stability.

[0043] 4. The fixed-angle jet mode is adopted, and the drone does not need to be equipped with a motor and a matching power and transmission system. This not only reduces its own load and increases its endurance, but also avoids random impacts and interference to the stability of the drone when actively adjusting the jet angle. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a principle flow chart of a UAV fixed-point spraying operation method based on the KLMS algorithm of the present invention;

[0045] Figure 2 Schematic diagram of the error value of the LMS algorithm under different step sizes;

[0046] Figure 3 Schematic diagram of the error value of the RLS algorithm under different step sizes;

[0047] Figure 4 Schematic diagram of the error value of the KLMS algorithm under different step sizes;

[0048] Figure 5 This is a simulation comparison chart of the LMS algorithm, RLS algorithm and KLMS algorithm;

[0049] Figure 6The spray point distribution diagrams obtained by using the LMS algorithm, RLS algorithm and KLMS algorithm respectively. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It is hereby stated that any directional terms such as "up," "down," "left," "right," "front," "back," "inside," and "outside" that appear or will appear herein are based solely on the accompanying drawings and are not intended to limit the present invention.

[0051] like Figure 1 As shown, this embodiment provides a UAV fixed-point spraying operation method based on the KLMS algorithm, comprising the following steps:

[0052] S1. The drone carrying the laser transmitter is launched into the air and performs laser marking to confirm the coordinates of the injection point;

[0053] S2. After confirming the coordinates of the jet landing point, the initial operating position of the UAV is calculated based on the preset parameters including the horizontal operating distance and jet speed;

[0054] S3: The UAV arrives at the initial operation position and begins spraying after being approved by the ground supervisor;

[0055] S4. During the operation, the KLMS-based landing point self-stabilization algorithm is executed to adjust the attitude and orientation of the UAV to ensure the stability of the jet landing point.

[0056] Specifically, this embodiment uses a separate water-carrying mode in which the drone is separated from the water tank. Specifically, the water tank is placed on the ground, with one end of the water pipe connected to the tank and the other end towed by the drone. This mode offers the advantage of reducing the drone's payload and ensuring optimal operation time. Furthermore, the interference from the water pipe on the drone can be assumed to be constant, facilitating algorithmic offset. However, when operating with the water tank, the drone's center of gravity shifts as the water level decreases, causing the drone's mathematical model to deviate from actual conditions and weakening its stability.

[0057] In addition, this embodiment adopts a spray mode with a fixed spray angle, so that the drone does not need to be equipped with a motor and a matching power and transmission system. This not only reduces its own load and increases its endurance, but also avoids random impacts and interference to the stability of the drone when actively adjusting the spray angle.

[0058] Specifically, in this embodiment, the KLMS-based landing point self-stabilization algorithm is executed to adjust the attitude and orientation of the UAV to ensure the stability of the injection landing point, specifically including:

[0059] Obtain the position and attitude angle information of the UAV under the influence of disturbance, and calculate the actual injection point d(n) based on the ejection. The calculation process includes:

[0060] A right-handed coordinate system is established with the laser point as the origin P0(0,0,0), and the Y-axis direction is along the overhead line. At a certain moment, the drone is located at P(x0,y0,z0), and its attitude angles are The jet velocity is v. Assuming that there is no air resistance during the jetting process, the motion of the jetted liquid in the air at a certain moment is oblique projectile motion.

[0061] According to the laws of motion, the displacement of water when it reaches the YOZ plane is calculated, thereby predicting the jet landing point at a certain moment:

[0062]

[0063] Among them, x0 is the X-axis coordinate, is the roll angle around the X axis, θ is the pitch angle around the Y axis, ψ is the yaw angle around the Z axis, and g is the gravity constant.

[0064] Then, the error e(n) between the actual injection landing point d(n) and the theoretical injection landing point x(n) is calculated;

[0065] Update the weight coefficient w(n) through the error e(n) and the convergence coefficient μ;

[0066] Apply the updated weight coefficient w(n) to the UAV dynamics model to change the UAV's power, thereby adjusting the UAV's attitude and orientation, and predicting the UAV's position at the next moment;

[0067] When the drone's position crosses the electronic fence at the next moment, the drone will be controlled to return to the electronic fence first; if the drone's position does not cross the electronic fence at the next moment, the drone will continue to predict the injection landing point and complete the next cycle iteration.

[0068] The electronic fence described in this embodiment is created with the initial operating position of the drone as the center.

[0069] Because the drone control in this embodiment focuses on stabilizing the spray point rather than the drone's own position and attitude, this means that while the drone's spray point remains focused on the target area and spraying operations are guaranteed, the drone's position and direction can be slightly deviated, thus achieving a multi-angle spraying effect. For operational effectiveness and safety distance considerations, the drone's deviation must be limited. Therefore, this embodiment limits the range of the drone's deviation during the self-stabilization control of the spraying operation, i.e., establishes an electronic fence.

[0070] Specifically, in order to combat the impact of random disturbances on the UAV's spraying operation, this embodiment adopts a control strategy based on the KLMS algorithm. The KLMS algorithm introduces the Gaussian kernel function on the basis of the LMS algorithm, mapping the high-dimensional original input data {x(1), x(2), ..., x(N)} (specifically the spray landing point) into a high-dimensional space. Realize the transformation of nonlinear problems into linear problems.

[0071] Assume there is a mapping X→H maps x(i) to The corresponding output is {y(1), y(2), …, y(N)}. The iterative process of the KLMS algorithm is:

[0072] w(0)=0

[0073]

[0074]

[0075] Where e(n) is the error, d(n) is the actual injection point, μ is the convergence coefficient, and w(n) is the weight coefficient vector when the iteration is performed to the nth time, which is recorded as:

[0076]

[0077] Let the kernel function be:

[0078]

[0079] Then there are estimated values:

[0080]

[0081] The output of the KMLS algorithm is independent of the weight coefficient. By establishing a kernel function, multiplying it with the corresponding error and then linearly combining them, the complete output of the filter is obtained, thereby realizing the dimensionality reduction processing of nonlinear filtering.

[0082] Specifically, in this embodiment, the UAV dynamics model is as follows:

[0083]

[0084] in, is the second-order derivative of any variable x in the x-axis direction, is the second-order derivative of any variable y in the y-axis direction, is the second-order derivative of any variable z in the z-axis direction, is the first-order derivative of the yaw angle, is the first derivative of the pitch angle, is the first derivative of the roll angle, They are the control quantities of the three directions of the angular velocity of the quadrotor drone, U1 is the virtual control quantity of lift, U2, U3, and U4 are the virtual control quantities of roll angle, pitch angle, and yaw angle respectively, J1, J2, and J3 are the control quantities of the three directions of moment of inertia respectively, l is the distance between the center of mass of the drone and the rotor motor, and m is the mass of the drone.

[0085] In order to test the control effect of the UAV jet landing point self-stabilization algorithm proposed in this embodiment, a simulation platform is built for verification calculation. First, the UAV state is initialized. It is assumed that the UAV obtains the operation target, and after the initial calculation, it starts and reaches the ideal preset position, facing the target. The current attitude angle is Then the attitude disturbance of the UAV is recorded as:

[0086] D=[d1 d2 d3] T +[η1 η2 η3] T

[0087] where [d1 d2 d3] T represents the random disturbance during flight, [η1 η2 η3] T represents the deviation caused by water pipe pulling and jet recoil. In the calculation, the recoil disturbance is set to a constant value and the windage is set to a step value during the action time to simplify the calculation. For ease of calculation, this embodiment assumes that the control signal output by the control system is always within the maximum driving force range of the drone, the sampling frequency is set to 100Hz, and the kernel function is selected as:

[0088]

[0089] To examine the ability to adjust to different disturbances, this example employed a small-amplitude disturbance between 0 and 0.5 seconds. A moderate-intensity disturbance, lasting 0.2 seconds, was introduced at 0.8 seconds, based on actual operation. Full-speed injection was then initiated at 1.5 seconds, representing a strong disturbance. To ensure greater generalizability, this example also employed adaptive controllers based on the LMS and RLS algorithms to compare their control performance.

[0090] First, we compare the control effects of the same algorithm under different parameters. The error values ​​of the three algorithms under different step sizes are as follows: Figure 2-Figure 4As shown. For the LMS algorithm, the algorithm response speed is slightly slower when the step size is 0.1, but the error is smaller in the stable state, the filter output is more accurate, and the relative effect is better; the overall effect of the RLS algorithm is better when the step size is 0.01; for the KLMS algorithm, when the step size is 0.01, the output error approaches zero, which means that the injection landing point in the theoretical calculation is always very close to the target landing point. However, such a relatively perfect virtual simulation result may be difficult to achieve in a real environment. Therefore, this embodiment selects a convergence curve with a step size of 0.1 for subsequent comparison.

[0091] The best results obtained by the three algorithms are as follows Figure 5 As shown in the figure, the control system based on the LMS adaptive filter has the largest steady-state error, the slowest response speed when subjected to sudden disturbances, and the largest error growth slope, indicating that the LMS algorithm has difficulty overcoming its inherent limitations. In comparison, the adaptive filter based on the RLS algorithm has significantly improved convergence speed and achieves the best regulation effect when the system is subjected to small-amplitude disturbances, but the control effect decreases as the disturbance increases.

[0092] The KLMS-based landing point self-stabilization system proposed in this example initially performed poorly, but after a period of basic learning and weight coefficient updates, control effectiveness improved significantly. Compared with the other two algorithms, the steady-state error was only one-quarter that of RLS control. Based on the above data and considerations, it can be concluded that the KLMS-based injection landing point self-stabilization system can improve the control effect of injection operations.

[0093] With the target (laser dot) as the center and 2cm as the radius, establish the effective operating range. Figure 6 (Left: LMS algorithm, Center: RLS algorithm, Right: KLMS algorithm) Shows the distribution of landing points 5 seconds after the start of spraying, measured using different algorithms. Because the pull of the pipeline applies a constant vertical force to the body, the overall distribution of the spray landing points is above the target point. In intuitive comparison, the KLMS algorithm has denser landing points and is more concentrated in the effective range, resulting in a better flushing effect. Statistical results also show that the stable algorithm based on the KLMS algorithm has an effective flushing rate of 93%, the RLS algorithm achieves an effective flushing rate of 84%, and the LMS algorithm can only achieve an effective flushing rate of 77%.

[0094] Although the description of the present disclosure has been quite detailed and particularly describes several embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but should be regarded as providing a broad possible interpretation of these claims in view of the prior art by reference to the appended claims, thereby effectively covering the intended scope of the present disclosure. In addition, the above description of the present disclosure is based on the embodiments foreseen by the inventors, which is intended to provide a useful description, and those non-substantial changes to the present disclosure that have not yet been foreseen may still represent equivalent changes to the present disclosure.

Claims

1. A UAV fixed-point spraying operation method based on the KLMS algorithm, characterized in that: The spray landing point is stabilized by controlling the drone, which specifically includes: The drone carries a laser transmitter and takes off into the air to perform laser marking to confirm the coordinates of the jet landing point; After confirming the coordinates of the jet landing point, the initial operating position of the UAV is calculated based on the preset parameters including the horizontal operating distance and jet speed. The UAV arrives at the initial operation position and begins spraying operations after being approved by the ground supervisor; During the operation, the KLMS-based landing point self-stabilization algorithm is executed to adjust the attitude and orientation of the UAV to ensure the stability of the jet landing point; The KLMS-based landing point self-stabilization algorithm is used to adjust the attitude and orientation of the drone to ensure the stability of the jet landing point. Specifically, the algorithm includes: Obtain the position and attitude angle information of the UAV under the influence of disturbance, and calculate the actual injection point d(n) based on the ejection; Calculate the error e(n) between the actual injection point d(n) and the theoretical injection point x(n); Update the weight coefficient w(n) through the error e(n) and the convergence coefficient μ; Apply the updated weight coefficient w(n) to the UAV dynamics model to change the UAV's power, thereby adjusting the UAV's attitude and orientation, and predicting the UAV's position at the next moment; The KLMS algorithm introduces the Gaussian kernel function based on the LMS algorithm to map the high-dimensional original input data {x(1), x(2),…, x(N)} into a high-dimensional space. Realize the transformation of nonlinear perturbation problems into linear problems; Assume there is a mapping Map x(i) to The corresponding output is {y(1), y(2), …, y(N)}. The iterative process of the KLMS algorithm is: w(0)＝0 Where e(n) is the error, d(n) is the actual injection point, μ is the convergence coefficient, and w(n) is the weight coefficient vector when the iteration is performed to the nth time, which is recorded as: Let the kernel function be: Then there are estimated values: The output of the KMLS algorithm is independent of the weight coefficient. By establishing a kernel function, multiplying it with the corresponding error and then linearly combining them, the complete output of the filter is obtained, thereby realizing the dimensionality reduction processing of nonlinear filtering.

2. The method for fixed-point spraying by an unmanned aerial vehicle based on the KLMS algorithm according to claim 1, characterized in that: A water-carrying mode in which the drone and the water tank are separated is adopted. Specifically, the water tank is placed on the ground, one end of the water pipe is connected to the water tank, and the other end is towed by the drone.

3. The method for fixed-point spraying of UAV based on KLMS algorithm according to claim 1, characterized in that: Adopts a spray pattern with a fixed spray angle.

4. The method for fixed-point spraying of an unmanned aerial vehicle based on the KLMS algorithm according to claim 1, characterized in that: When the next moment the drone's position crosses the electronic fence, the drone will be controlled to return to the electronic fence first; If the drone's position does not cross the electronic fence at the next moment, the drone will continue to predict the jet landing point and complete the next cycle iteration.

5. The method for fixed-point spraying of an unmanned aerial vehicle based on the KLMS algorithm according to claim 4, characterized in that: The electronic fence is created with the initial operating position of the drone as the center.

6. The method for fixed-point spraying by an unmanned aerial vehicle based on the KLMS algorithm according to claim 1, characterized in that: Calculates the spray drop point, including: A right-handed coordinate system is established with the laser point as the origin P0(0,0,0), and the Y-axis direction is along the overhead line. At a certain moment, the drone is located at P(x0,y0,z0), and its attitude angles are The jet velocity is v. Assuming that there is no air resistance during the jetting process, the motion of the jetted liquid in the air at a certain moment is oblique projectile motion. According to the laws of motion, the displacement of water when it reaches the YOZ plane is calculated, thereby predicting the jet landing point at a certain moment: Among them, x0 is the X-axis coordinate, is the roll angle around the X axis, θ is the pitch angle around the Y axis, ψ is the yaw angle around the Z axis, and g is the gravity constant.

7. The method for fixed-point spraying by an unmanned aerial vehicle based on the KLMS algorithm according to claim 1, characterized in that: The UAV dynamics model is as follows; in, is the second-order derivative of any variable x in the x-axis direction, is the second-order derivative of any variable y in the y-axis direction, is the second-order derivative of any variable z in the z-axis direction, is the first-order derivative of the yaw angle, is the first derivative of the pitch angle, is the first derivative of the roll angle, They are the control quantities of the three directions of the angular velocity of the quadrotor drone, U1 is the virtual control quantity of lift, U2, U3, and U4 are the virtual control quantities of roll angle, pitch angle, and yaw angle respectively, J1, J2, and J3 are the control quantities of the three directions of moment of inertia respectively, l is the distance between the center of mass of the drone and the rotor motor, and m is the mass of the drone.

Images