High-speed boat path tracking control method based on improved nonlinear speed observer

Abstract

The invention provides a high-speed boat path tracking control method based on an improved nonlinear speed observer, and the method comprises the steps: constructing a kinetic model and a kinematic model of an under-actuated water surface high-speed unmanned boat, designing a thrust saturation limiting function based on an ISRU, and analyzing a path tracking error and a control target; in a feedback function of an original nonlinear speed observer, fusing a hyperbolic tangent feedback mechanism of an error self-adaptive smoothing parameter, replacing a fixed gradient parameter with a parameter dynamically adjusted along with an observation error, substituting the parameter into a new feedback function, and reconstructing a nonlinear observer model; in the self-adaptive dynamic foresight sight guidance law, performing multi-state fusion improvement on a foresight distance, a navigational speed guidance law and a course guidance law; and improving a state estimator and a weight adaptive law of the RBF neural network. A dynamic error gain coefficient is introduced into a state estimator, a dynamic attenuation coefficient is introduced into a weight adaptive law, and the problems of model uncertainty and environmental interference in under-actuated ship trajectory tracking are effectively solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of high-speed boat path tracking control, and particularly relates to a high-speed boat path tracking control method based on an improved nonlinear velocity observer. BACKGROUND

[0002] Path tracking technology is an important guarantee for the safety, autonomy, precision and rapidity of unmanned boats in completing various tasks. The control goal of path tracking technology is to design a controller to enable the unmanned boat to accurately track and remain on a desired path in the tracking space independent of time. High-speed unmanned boats have characteristics such as nonlinearity, large inertia and time lag in the path tracking process. These characteristics make high-speed unmanned boats extremely susceptible to disturbances from complex environmental factors such as sea waves, wind and ocean currents. Therefore, designing a path tracking controller with strong adaptive ability, high robustness, strong anti-interference ability and high real-time performance is of important theoretical significance and practical value for the task adaptability and debugging safety of high-speed unmanned boats, and is a key technical support for promoting marine scientific and technological innovation and building a strong marine country. SUMMARY

[0003] The application aims to solve the problems of control precision loss, weak parameter adaptive ability of the guidance method and easy drift angle caused by environmental interference in the existing path tracking technology. The application provides a high-speed unmanned boat path tracking control method based on an improved nonlinear velocity observer.

[0004] The application provides a high-speed boat path tracking control method based on an improved nonlinear velocity observer, which comprises the following steps:

[0005] Step 1: constructing a mathematical model of a high-speed boat, designing a thrust saturation limiting function, and analyzing path tracking errors and control goals; the path tracking errors include lateral errors, lateral error change rates and speed tracking errors;

[0006] Step 2: in the feedback function of the original nonlinear velocity observer, according to the path tracking errors of step 1, a nonlinear feedback function is designed by utilizing the nonlinear smoothing property of the hyperbolic tangent function, a dynamic gradient parameter dynamically adjusted according to the observation error is used to replace the fixed gradient parameter, the nonlinear feedback function is substituted, and a nonlinear velocity observer model is reconstructed, wherein the nonlinear velocity observer model outputs a longitudinal velocity;

[0007] Step 3: in the adaptive dynamic look-ahead line-of-sight guidance law, the look-ahead distance, the speed guidance law and the heading guidance law are improved in multiple states, including: the longitudinal velocity output by step 2 and the lateral error change rate of step 1 are introduced into the look-ahead distance, the speed tracking error analyzed in step 1 is introduced into the speed guidance law, and the heading guidance law is combined with the lateral error and the error change rate of step 1; the expected heading angle and the expected combined speed are obtained through the speed guidance law and the heading guidance law;

[0008] Step 4: Based on the path tracking error of step 1, the dynamic error gain coefficient is designed, and the state estimator and weight adaptive law of the RBF neural network are improved, including: the state estimator introduces the dynamic error gain coefficient, the weight adaptive law introduces the dynamic attenuation coefficient, after the control quantity is generated through the backstepping method, the expected heading angle and the expected combined speed output by step 3 are referenced, and finally the thrust saturation limiting function designed in step 1 is compensated and processed, and then output to the actuator to control the high-speed boat to run until the last expected path point is tracked, and the tracking control process is completed.

[0009] Advantages: Compared with the prior art, the present application has the following advantages:

[0010] (1) By adaptively adjusting the dynamic gradient parameter according to the observation error, the estimation lag or high-frequency oscillation problem of the fixed parameter when the error suddenly changes is avoided, the speed estimation accuracy is improved, and more real-time and reliable speed information is provided for subsequent guidance and control;

[0011] (2) Compared with the original ADFLOS design optimization, the multi-state fusion ADFLOS design proposed in the present application can reduce the lateral deviation and speed oscillation amplitude, reduce the heading tracking error, ensure that the high-speed turns in advance, and still maintain smooth tracking when switching between complex curved paths, thereby improving the overall tracking accuracy;

[0012] (3) By introducing the dynamic attenuation coefficient, the weight update is matched with the error state, the neural network approximation of unknown items such as hydrodynamic parameter perturbation is more accurate, and the influence of model uncertainty on the control signal is reduced;

[0013] (4) The present application optimizes the state estimation through the dynamic error gain, cooperates with the segmented smooth processing of saturation compensation, reduces the thrust oscillation amplitude, avoids the control performance decline caused by actuator saturation, and improves the stability of the system under complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 It is a USV path tracking schematic diagram based on LOS guidance of the present application;

[0015] Figure 2 It is a path tracking trajectory comparison of three kinds of guidance control strategies of the present application;

[0016] Figure 3 It is an RBF neural network structure schematic diagram of the present application;

[0017] Figure 4 It is a neural network adaptive robust output feedback system block diagram of the adaptive dynamic forward-looking line-of-sight guidance of the present application. DETAILED DESCRIPTION

[0018] The technical features of the present application will be further clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described examples are only part of the embodiments of the present application, but not all the embodiments. Based on the examples in the present specification, all other embodiments obtained by those of ordinary skill in the art without creative work should fall within the scope of protection of the present specification.

[0019] The embodiment of the present application provides a high-speed boat path tracking control based on a nonlinear speed observer, mainly comprising the following steps:

[0020] Step 1: Constructing a dynamic model and a kinematic model of an underactuated high-speed unmanned surface vehicle, designing a thrust saturation limiting function based on ISRU, and analyzing path tracking error and control target. Specifically, it includes:

[0021] Taking an underactuated high-speed unmanned surface vehicle with a single propeller and a single rudder as a carrier, considering the characteristics of high-speed unmanned surface vehicles such as nonlinearity, strong coupling and large inertia, ignoring the heave, roll and pitch motion of the unmanned surface vehicle, and only considering the horizontal motion of three degrees of freedom of surge, sway and yaw, the simplified nonlinear mathematical model of the underactuated high-speed vehicle is represented as:

[0022]

[0023] In the formula, denotes the position of the unmanned surface vehicle in the inertial coordinate system and the heading angle , is the longitudinal velocity, lateral velocity and yaw angular velocity of the unmanned surface vehicle in the ship coordinate system, is the conversion matrix between the inertial coordinate system and the ship coordinate system, and the relationship is as follows:

[0024]

[0025] is the mass inertia matrix of the unmanned surface vehicle, is the Coriolis centripetal force matrix, which is represented as follows:

[0026]

[0027] is the fluid damping coefficient matrix, is the longitudinal thrust and the steering torque . The underactuated high-speed vehicle in the present application only contains one main propeller and one rudder, without lateral control input. is an unknown time-varying disturbance caused by environmental factors such as wind, wave and current, is the upper bound of the disturbance. At the same time, the change of the sailing speed of the high-speed vehicle during tracking will cause the perturbation of the hydrodynamic parameters and further produce the unknown items of the model .

[0028] Combining the simplified nonlinear mathematical model of the underactuated high-speed boat and the model uncertainty terms, the nonlinear mathematical model for the motion control of the high-speed boat can be obtained:

[0029]

[0030] During actual navigation of the high-speed boat, considering the speed limit of the USV propeller and the saturation of the rudder angle, constraints are imposed on the longitudinal thrust generated by the main propeller and the steering torque generated by the rudder, and a saturation constraint function is constructed:

[0031]

[0032] In the formula, As the initial control input, To control the limit values ​​of the input. The control input is transformed by the input restricted function. To address the potential abrupt changes in thrust output caused by traditional sign functions, an inverse square root linear unit is introduced to ensure that the thrust output remains unsaturated throughout the entire input domain. This guarantees the smoothness of the thrust output and avoids the thrust saturation problem caused by slightly larger control inputs. While ensuring the operational safety of the high-speed boat path tracking control system, this also improves the maneuverability and flexibility of the high-speed boat during tracking. The expression is:

[0033]

[0034] In the formula, for The gradient design parameters, and .

[0035] The ship's coordinate system has its origin at the high-speed vessel's center of gravity, with the bow direction as the X-axis and the direction perpendicular to the bow as the Y-axis. The desired tracking path consists of a series of parameterized path points. It is connected together, among which These are time-independent path parameters. Representing path points The rotation angle of the tangential coordinate system relative to the inertial coordinate system at point , where , . and These represent longitudinal tracking error and lateral tracking error, respectively. and These represent the longitudinal and lateral speeds of the high-speed boat, respectively. This indicates the actual total speed of the high-speed boat. This indicates the drift angle during the high-speed boat's navigation. denotes the actual yaw angle of the high-speed boat, denotes the desired course angle after LOS guidance, denotes the forward-looking distance of the LOS guidance. The final high-speed boat position tracking error is expressed as:

[0036]

[0037] Differentiating the above equation, and respectively, the USV position tracking error dynamic equation is obtained as:

[0038]

[0039] It can be seen that the kinematic control objective of high-speed boat path tracking is to construct a nonlinear velocity observer to estimate the high-speed boat velocity information in the presence of external environmental disturbances, model parameter perturbations and input constraints, and in the case of unobservable velocity vector, and then generate the desired speed and heading angle through the guidance law, and further design the speed and heading control law to make the USV track the desired path.

[0040] Step 2: In the feedback function of the original nonlinear velocity observer, the error adaptive smoothing parameter designed by step 1 path tracking error is fused into the tangent hyperbolic feedback mechanism, and the fixed gradient parameter is replaced by the parameter dynamically adjusted according to the observation error, and the new feedback function is substituted into the nonlinear observer model. Specifically, it includes:

[0041] Step (2.1): Assuming that the longitudinal coordinate, transverse coordinate and bow angle of the high-speed boat in the inertial coordinate system are measurable, and the longitudinal velocity, transverse velocity and yaw angle velocity are unmeasurable. The Coriolis centripetal force matrix is constructed as follows: , and the state transition matrix is constructed as follows:

[0042]

[0043] In the equation, is a generalized variable matrix.

[0044] Further differentiation gives:

[0045]

[0046] In order to make , select as:

[0047]

[0048] In the equation,

[0049]

[0050] Step (2.2): is obtained by solving the equations simultaneously:

[0051]

[0052] where, .

[0053] Step (2.3): The nonlinear observer is designed as follows:

[0054]

[0055] where, , is , the observation value of , is the gain matrix, is the position observation error of the nonlinear observer. , is the nonlinear feedback function, and .

[0056] The nonlinear function is designed as follows:

[0057]

[0058] where, the dynamic gradient parameter and are defined as:

[0059]

[0060] where, is the basic gradient parameter, is the gradient adjustment amplitude, is the error sensitivity coefficient, is the observation error.

[0061] Step (2.4): According to the high-speed boat position tracking error, the nonlinear feedback function is designed by using the nonlinear smoothing characteristics of the hyperbolic tangent function, which can effectively avoid the problem of control quantity oversaturation caused by too large observation error. At the same time, through the dynamic gradient parameters and , the problem of insufficient control quantity when the observation error is small can be solved, and high-frequency oscillation in the observation process can be avoided. The final nonlinear observer model is represented as:

[0062]

[0063] where, is the speed estimate obtained by the observer, , , are the observation values of the longitudinal speed, lateral speed and yaw angle speed, and satisfy , , .

[0064] Step (2.5): Define the velocity observation error as , the observation error of the nonlinear velocity observer is , combining step (2.1) and step (2.2) can be obtained:

[0065]

[0066] In the formula, is the position observation vector, is the observation error vector.

[0067] The Lyapunov function is constructed as follows:

[0068]

[0069] Take the derivative and substitute it into the above formula:

[0070]

[0071] In the formula, , , , , is a positive definite matrix, and satisfies:

[0072]

[0073] Step (2.6): From the condition defined by the nonlinear feedback function and step (2.3), we can get , and finally according to step (2.5) we can get:

[0074]

[0075] Therefore, the observation error of the nonlinear velocity observer is asymptotically stable, and ultimately converges to zero. The mathematical model of the underactuated high-speed ship motion estimated by the nonlinear velocity observer is transformed into the following form:

[0076]

[0077] Step 3: In order to solve the problems of slow tracking convergence speed, poor dynamic adaptability and weak adaptive ability of the heading in the traditional line-of-sight guidance method, the multi-state fusion mechanism is introduced to improve the look-ahead distance, speed guidance law and heading guidance law in the adaptive dynamic look-ahead line-of-sight guidance law (ADFLOS). The longitudinal velocity output in step 2 and the lateral error rate in step 1 are introduced into the look-ahead distance, the speed guidance law is introduced into the speed tracking error analyzed in step 1, and the heading guidance law is combined with the lateral error and error rate in step 1. The expected heading angle and expected combined speed are obtained through the speed and heading guidance law. Specifically, it includes:

[0078] Step (3.1): The dynamic multi-state fusion mechanism is introduced to optimize the expected heading, so that the USV can dynamically adjust the look-ahead distance according to the tracking error, speed and error rate, and simultaneously improve the adaptive ability of the expected heading. The speed, heading and virtual target guidance law in ADFLOS is designed as follows:

[0079]

[0080] In the formula, is the expected speed obtained by the speed guidance law, is the expected speed coefficient ( ), is the speed tracking coefficient, is the reference speed, is the gradient parameter of the error rate, is the lateral error rate, is the look-ahead distance.

[0081] Step (3.2): In the traditional line-of-sight method, Generally, it is 3-5 times the length of the ship. For high-speed boats, fixed look-ahead distance will cause large lateral deviation and slow convergence speed when tracking curves or switching paths. To improve this drawback, the tracking error of the unmanned ship is combined with the look-ahead distance. When the tracking error is large, the look-ahead distance is reduced, and the USV can converge to the expected path at a faster speed. When the tracking error is small, the look-ahead distance increases, and the convergence speed of the USV slows down, which can smoothly track the expected path and avoid the unmanned ship oscillating along the expected path at high speed. The look-ahead distance is a multi-state fusion adaptive form, which combines the longitudinal velocity and the lateral error rate in step (1.4) and is dynamically adjusted as follows:

[0082]

[0083] In the formula, is a positive design parameter, is the length of the unmanned ship, is the longitudinal velocity adaptation coefficient, is the error rate response coefficient, which increases the look-ahead distance at high speed and enhances the response when the error changes suddenly.

[0084] Step (3.3): Introduce the hyperbolic tangent function related to the lateral error in the heading guidance law to improve the adaptive ability of the expected heading, is the positive gradient design parameter of the hyperbolic tangent function, is the drift angle compensation designed by combining the expected speed with the lateral velocity estimated by the nonlinear observer.

[0085] Step (3.4): Obtain the expected heading angle in step (3.1), is the reference speed of the virtual target point obtained by the virtual target guidance law, which is the control input that promotes the convergence of the longitudinal tracking error . Where, is the longitudinal error coefficient, is the expected combined speed. Substitute in step (1.4) into step (2.3) to obtain the update law of the path parameter as follows:

[0086]

[0087] Step 4: Based on the error design in step 1, the dynamic error gain coefficient is improved. The state estimator of the RBF neural network and the weight adaptive law are improved. The state estimator introduces the dynamic error gain coefficient, and the weight adaptive law introduces the dynamic decay coefficient. After generating the control quantity by the backstepping method, the expected heading angle and the expected combined speed output by step 3 are referenced, and finally the ISRU thrust saturation limiting function compensation processing designed in step 1 is output to the actuator to control the unmanned surface vehicle to run until the last expected path point is tracked, and the tracking control process is completed. Specifically, it includes:

[0088] Step (4.1): Combine the expected heading angle obtained by the high-speed boat dynamics model in step (2.6) and the dynamic line-of-sight guidance method to design the USV heading control law as follows:

[0089]

[0090] where, is the positive design parameter of the controller, is the heading tracking error. In the case where the USV speed is difficult to measure, the rudder angular velocity error can be defined as where is the virtual control law of the turning moment. is the upper bound of external disturbance the estimated value of , the unknown disturbance upper bound is estimated by the disturbance adaptive law, which improves the robustness of the control system, and the adaptive law is designed as follows:

[0091]

[0092] wherein , is a positive design parameter, is the priori estimated value of .

[0093] Step (4.2): The unknown term is approximated by the improved radial basis neural network. According to the universal approximation property of the RBF neural network, the output expression of the RBF neural network is designed as follows:

[0094]

[0095] wherein is the input vector after speed estimation, is the radial basis function vector, and the expression of the basis function is:

[0096]

[0097] wherein is the center point vector value of the Gaussian function, and the dimension is the same as the input vector , is the width of the Gaussian function, is the number of nodes of the neural network hidden layer. is the approximation error, and satisfies , is the error bound value. represents the weight vector from the hidden layer to the output layer, and when the approximation error is the smallest, is the ideal weight vector, and in the actual engineering application, the estimated value of is used to design the corresponding control law.

[0098] Step (4.3): The estimation error is obtained by introducing the state estimator, and then the adaptive law of the neural network is designed, and the state estimator is designed as follows:

[0099]

[0100] wherein is the estimated value of , is the estimation error of the state estimator to , is the dynamic error gain coefficient:

[0101]

[0102] where is the basic error gain, is the gain adjustment amplitude, is the error sensitive coefficient, which realizes the dynamic adaptation of reducing the gain to accelerate the convergence when the error is large and increasing the gain to suppress the oscillation when the error is small.

[0103] Step (4.4): Based on the estimation error of the dynamic state estimator output , the weight update law is designed to adaptively adjust the error, and the dynamic attenuation coefficient is introduced, the expression is:

[0104]

[0105] where, is the positive design matrix in the adaptive law, is defined as:

[0106]

[0107] where, is the basic attenuation coefficient, is the attenuation adjustment amplitude, is the attenuation sensitive coefficient, which makes the weight update rate adapt to the system dynamic error, accelerates the weight convergence when the error is large, suppresses the weight oscillation when the error is small, and improves the approximation accuracy of unknown terms.

[0108] Step (4.5): The adaptive law in step (4.4) does not depend on the tracking error, but obtains the ideal weight vector according to the estimation error of the state estimator, which improves the dynamic performance of the entire path tracking control system. In order to reduce the influence of input constraints, the auxiliary observation system is introduced to combine the thrust saturation constraint function constructed in step 1 to compensate for the yaw angle velocity error , the auxiliary observation system is designed as follows:

[0109]

[0110] where, , , is the constructed thrust saturation constraint function.

[0111] Step (4.6): When the control law is designed by using backstepping method, the derivative of the virtual control law contains the second derivative of the desired heading angle Therefore, in order to further improve the efficiency of solving the control law, a second-order filter is designed to obtain the derivative estimate of the desired heading angle. The specific design is as follows:

[0112]

[0113] In the formula, , Let be the state variable of the second-order filter. , The filter design parameters are as follows. The first and second derivatives of the desired heading angle after filtering by the second-order filter are respectively... and This leads to the replacement of the derivative of the virtual control law. In and .

[0114] Step (4.7): Combining the dynamics model of the high-speed boat under the condition of unmeasurable speed and the desired speed, the speed control law of the USV is designed as follows:

[0115]

[0116] in, , For positive design parameters, For longitudinal velocity tracking error, External interference with the upper realm The estimated value, The perturbation adaptive law design is as follows:

[0117]

[0118] In the above formula, for The prior estimate, This is the virtual control law after DSC filtering. This represents the ideal weight vector that minimizes the approximation error of the neural network. This is the estimation term for the uncertain part of the model in the vertical direction by the neural network. The output expression, state estimator, and weight adaptive law of the neural network are designed as follows:

[0119]

[0120] in, To approximate the error, for The estimated value, The estimation error of the state estimator, This is the error gain coefficient. This is the positive definite design matrix in the adaptive law. Designing positive parameters.

[0121] Step (4.8): Designing an auxiliary compensation system according to the thrust saturation limiting function to compensate for the velocity tracking error , and thus reduce the impact of the thruster input limitation. The auxiliary compensation system is designed as follows:

[0122]

[0123] where , .​

Claims

1. A path tracking control method for high-speed boats based on an improved nonlinear velocity observer, characterized in that: Includes the following steps: Step 1: Construct a mathematical model of the high-speed boat, design a thrust saturation constraint function, and analyze the path tracking error and control target; the path tracking error includes lateral error, lateral error rate of change, and speed tracking error; Step 2: In the feedback function of the original nonlinear velocity observer, based on the path tracking error in Step 1, a nonlinear feedback function is designed using the nonlinear smoothing characteristics of the hyperbolic tangent function. The fixed gradient parameter is replaced with a dynamic gradient parameter that is dynamically adjusted according to the observation error. The designed nonlinear feedback function is substituted into the nonlinear feedback function to reconstruct the nonlinear velocity observer model. The nonlinear velocity observer model outputs the longitudinal velocity. Step 3: In the adaptive dynamic forward-looking line-of-sight guidance law, multi-state fusion improvement is performed on the forward-looking range, speed guidance law, and heading guidance law, including: incorporating the longitudinal velocity output in Step 2 and the lateral error change rate in Step 1 into the forward-looking range; incorporating the speed tracking error analyzed in Step 1 into the speed guidance law; and combining the lateral error and error change rate in Step 1 into the heading guidance law; the desired heading angle and desired resultant velocity are obtained through the speed guidance law and the heading guidance law. Step 4: Based on the path tracking error in Step 1, design a dynamic error gain coefficient to improve the state estimator and weight adaptive law of the RBF neural network. This includes: introducing a dynamic error gain coefficient into the state estimator and a dynamic attenuation coefficient into the weight adaptive law. After generating the control quantity through the backstepping method, refer to the desired heading angle and desired resultant velocity output in Step 3. Finally, after being compensated by the thrust saturation limiting function designed in Step 1, the output is sent to the actuator to control the high-speed boat until it tracks the last desired path point, thus completing the tracking control process.

2. The path tracking control method for a high-speed boat based on an improved nonlinear velocity observer according to claim 1, characterized in that: The mathematical model for constructing a high-speed boat is described, including the design of a thrust saturation constraint function and the analysis of path tracking error and control objectives. The path tracking error includes lateral error, lateral error rate of change, and speed tracking error. Specifically, it includes: Considering only the horizontal plane motion with three degrees of freedom—swell, roll, and yaw—the simplified nonlinear mathematical model of the underactuated high-speed boat is expressed as follows: In the formula, Indicates the position of the high-speed boat in the inertial coordinate system. and heading angle , Let these be the longitudinal velocity, lateral velocity, and bow roll rate in the ship's coordinate system. Here is the transformation matrix between the inertial coordinate system and the coordinate system of the high-speed boat, with the following relationship: Here is the mass inertia matrix of the high-speed boat. The Coriolis centripetal force matrix is ​​represented as follows: This is the fluid damping coefficient matrix. For longitudinal thrust and steering torque ; Combining the simplified nonlinear mathematical model of the underactuated high-speed boat and the model uncertainty terms The nonlinear mathematical model for the motion control of high-speed boats is obtained as follows: To constrain and limit the longitudinal thrust generated by the main propulsion system and the steering torque generated by the rudder of the high-speed boat, a saturation constraint function is constructed: In the formula, As the initial control input, To control the input limit values, The expression is: In the formula, for The gradient design parameters, and ; The coordinate system of the high-speed boat is based on its center of gravity, with the bow direction as the X-axis and the direction perpendicular to the bow as the Y-axis. The desired tracking path consists of a series of parameterized path points. It is connected together, among which These are time-independent path parameters. Representing path points The rotation angle of the tangential coordinate system relative to the inertial coordinate system at point , where , , and These represent longitudinal tracking error and lateral tracking error, respectively. and These represent the longitudinal and lateral velocities of the high-speed boat, respectively. This indicates the actual total speed of the high-speed boat. This indicates the drift angle during the high-speed boat's navigation. This indicates the actual bow roll angle of the high-speed boat. This represents the desired heading angle after line-of-sight guidance. Indicates the forward look-ahead distance for LOS guidance; The path tracking error is expressed as: For the formula and Taking the derivatives separately, we obtain the dynamic equation for the high-speed boat path tracking error as follows: 。 3. The path tracking control method for a high-speed boat based on an improved nonlinear velocity observer according to claim 2, characterized in that: In the feedback function of the original nonlinear velocity observer, based on the path tracking error in step 1, a nonlinear feedback function is designed using the nonlinear smoothing characteristics of the hyperbolic tangent function. The fixed gradient parameters are replaced with dynamic gradient parameters that are dynamically adjusted according to the observation error. The designed nonlinear feedback function is substituted into the nonlinear feedback function to reconstruct the nonlinear velocity observer model, which outputs the longitudinal velocity. Specifically, it includes: Step (2.1): From the Coriolis centripetal force matrix The state transition matrix is ​​constructed as follows: In the formula, It is a generalized variable matrix; Further differentiation yields: In order to make Select for: In the formula: Step (2.2): Simultaneous equations yield: in, ; Step (2.3): The nonlinear velocity observer is designed as follows: In the formula, , for , The observed values, , Here is the gain matrix. The position observation error of the nonlinear velocity observer, , It is a nonlinear feedback function, and ; The nonlinear function is designed as follows: Among them, dynamic gradient parameters and Defined as: In the formula, Based on the basic gradient parameters, For gradient adjustment amplitude, For error sensitivity coefficient, This is the observation error; Step (2.4): Based on the high-speed boat path tracking error, a nonlinear feedback function is designed using the nonlinear smoothing characteristic of the hyperbolic tangent function, and the dynamic gradient parameters are used to... and Finally, the nonlinear velocity observer model is obtained, which is expressed as: in, The velocity estimate obtained from the nonlinear velocity observer model. , , The observed values ​​are the longitudinal velocity, lateral velocity, and bow roll angular velocity, and satisfy the following conditions: , , .

4. The path tracking control method for a high-speed boat based on an improved nonlinear velocity observer according to claim 3, characterized in that: The adaptive dynamic forward-looking line-of-sight guidance law described above involves multi-state fusion improvement of the forward-looking range, speed guidance law, and heading guidance law. This includes: incorporating the longitudinal velocity output in step 2 and the lateral error rate of change in step 1 into the forward-looking range guidance law; incorporating the speed tracking error analyzed in step 1 into the speed guidance law; and combining the lateral error and error rate of change in step 1 into the heading guidance law. The desired heading angle and desired resultant velocity are obtained through the speed guidance law and the heading guidance law. Specifically, this includes: Step (3.1): In the adaptive dynamic forward-looking line-of-sight guidance law, design the guidance laws for speed, heading, and virtual target, expressed as: In the formula, The desired speed obtained from the speed guidance law. For the desired speed coefficient, For speed tracking coefficient, For reference speed, Let be the gradient parameter of the rate of change of error. This represents the rate of change of the lateral error. Forward sight distance; The reference velocity of the virtual target point is obtained from the virtual target guidance law. This is the longitudinal error coefficient. For the desired resultant velocity; Step (3.2): Combine the longitudinal velocity from step 1 and the rate of change of lateral error Dynamically adjust forward viewing distance , is represented as: In the formula, For positive design parameters, Captain of the high-speed boat, For longitudinal velocity adaptation coefficient, The error rate of change response coefficient; Step (3.3): Design drift angle compensation using the lateral velocity estimated by the nonlinear velocity observer combined with the desired speed, expressed as: ; Step (3.4): Substituting into step (2.3), we obtain the path parameters. The update law is as follows: 。 5. The path tracking control method for a high-speed boat based on an improved nonlinear velocity observer according to claim 4, characterized in that: The aforementioned design of a dynamic error gain coefficient based on the path tracking error in step 1 improves the state estimator and weight adaptive law of the RBF neural network. This includes: introducing a dynamic error gain coefficient into the state estimator and a dynamic attenuation coefficient into the weight adaptive law. After generating the control quantity through backstepping, referencing the desired heading angle and desired resultant velocity output in step 3, and finally compensating through the thrust saturation constraint function designed in step 1, the output is sent to the actuator to control the high-speed boat's operation until it tracks the last desired path point, completing the tracking control process. Specifically, this includes: Step (4.1): Combining the mathematical model of the high-speed boat with the desired heading angle obtained by the dynamic line-of-sight guidance method, the USV heading control law is designed as follows: In the formula, For the positive design parameters of the controller, For heading tracking error; Based on the estimates from the nonlinear velocity observer, the bow roll angular velocity error is defined as... ,in This is a virtual control law for steering torque; External interference with the upper realm The estimated value, An adaptive law is used to estimate the upper bound of unknown external disturbances. The adaptive law is designed as follows: in, , For positive design parameters, for The prior estimate; Step (4.2): Design the following RBF neural network output expression: In the formula, The input vector is the one after velocity estimation. Let be the radial basis function vector, and the expression for the basis functions is: in, The center point vector values ​​of the Gaussian function, dimension and input vector same, The width of the Gaussian function. The number of nodes in the hidden layer of the neural network. To approximate the error, and satisfy... , This is the error limit value; This represents the weight vector from the hidden layer to the output layer, where the approximation error is minimized. The ideal weight vector; Step (4.3): Obtain the estimation error by introducing a state estimator, and then design the adaptive law of the neural network. The state estimator is designed as follows: In the formula, for The estimated value, For state estimator pairs The estimation error, For dynamic error gain coefficient: in Based on the basic error gain, For gain adjustment range, This is the error sensitivity coefficient; Step (4.4): Estimation error based on the output of the dynamic state estimator A weight update law that adaptively adjusts with error is designed, and a dynamic decay coefficient is introduced. The expression is: in, This is the positive design matrix in the adaptive law. Defined as: in, Based on the attenuation coefficient, To attenuate the adjustment amplitude, This is the attenuation sensitivity coefficient; Step (4.5): Introduce an auxiliary observation system and combine it with the thrust saturation constraint function constructed in step 1 to compensate for the bow roll rate error. The auxiliary observation system is designed as follows: in, , , The thrust saturation constraint function is constructed. Step (4.6): When designing the control law using the backstepping method, the derivative of the virtual control law... The second derivative containing the desired heading angle A second-order filter is designed to obtain the derivative estimate of the desired heading angle. The specific design is as follows: In the formula, , Let be the state variable of the second-order filter. , The filter design parameters are as follows; the first and second derivatives of the desired heading angle after filtering by the second-order filter are respectively and The derivative of the virtual control law is replaced. In and ; Step (4.7): The speed control law for the high-speed boat is designed as follows: in, , For positive design parameters, For longitudinal velocity tracking error, External interference with the upper realm The estimated value, The perturbation adaptive law design is as follows: In the above formula, for The prior estimate, This is the filtered virtual control law. This represents the ideal weight vector that minimizes the approximation error of the neural network. The neural network estimates the uncertainties in the vertical direction of the model. The output expression, state estimator, and adaptive weight law of the neural network are designed as follows: in, To approximate the error, for The estimated value, The estimation error of the state estimator, This is the error gain coefficient. This is the positive definite design matrix in the adaptive law. Positive design parameters; Step (4.8): Design the auxiliary compensation system based on the thrust saturation limit function. To compensate for speed tracking error The auxiliary compensation system is designed as follows: In the formula, , .

Images