ADRC-based navigation planning and control method for autonomous sightseeing ship

Abstract

The application relates to the technical field of ship adjustment, and discloses an automatic driving sightseeing ship navigation planning and control method based on ADRC, which comprises the following steps: acquiring sightseeing ship pose feedback and front wave disturbance data; mapping the disturbance data into characteristic sampling points and fitting to generate a disturbance space-time evolution envelope; determining a gradient gain factor according to the principal curvature of the evolution envelope at a prediction point; setting the observation bandwidth of an extended state observer by using the gradient gain factor, estimating a lumped disturbance quantity containing environmental disturbance and unmodeled dynamics; and generating an initial adjustment quantity according to a path deviation and outputting a thrust instruction in combination with the lumped disturbance quantity. The application actively adjusts the observation bandwidth by using differential geometry characteristics, changes disturbance perception from lagging response to early prediction, effectively suppresses observation logic divergence caused by physical limiting of a power system, and guarantees path tracking fidelity in complex water areas.

Description

Technical Field

[0001] This invention belongs to the field of ship regulation technology, and in particular relates to a navigation planning and control method for autonomous sightseeing boats based on ADRC. Background Technology

[0002] Currently, navigation missions of mobile surface platforms mainly rely on attitude control systems, which are used for closed-loop feedback regulation of the ship's motion. The mainstream solutions mostly adopt proportional-integral-derivative control or predictive control methods based on physical models. By collecting inertial navigation data and desired route data, the rudder angle and speed adjustment commands are calculated. Sightseeing boats are typical underactuated systems. When navigating in unstructured dynamic waters, they are strongly affected by time-varying nonlinear hydrodynamic damping and external random disturbances such as wind, waves, and currents. Existing regulation strategies, in order to maintain tracking accuracy, usually select high control gain or improve system response by refining dynamic modeling parameters. However, due to the non-stationary characteristics of the water surface environment, high gain configurations often cause the actuators to frequently exceed physical limits, resulting in command overshoot. The determination of model parameters is limited by sensor accuracy and environmental adaptability.

[0003] To address the aforementioned interference suppression requirements, the industry has attempted to introduce an active disturbance rejection control architecture, treating internal coupling and external interference as a total disturbance, which is then estimated in real time by an extended state observer. Analysis reveals that the observer's bandwidth selection is constrained by the relationship between sensor noise and the accuracy of total disturbance estimation. While simply increasing the bandwidth enhances the ability to capture high-frequency disturbances, it also introduces an amplification effect from sensor noise. Conventional variable bandwidth strategies generally compress multidimensional environmental disturbances into a one-dimensional time-series scalar, ignoring the topological asymmetry of the physical disturbance field in space. This results in the system being unable to pre-adjust adjustment parameters when facing sudden surges with steep morphology, leading to transient yaw. These problems partly stem from hardware limitations, such as the specific hull shape applicable to certain water areas or the inherent limitations of the propeller's physical structure, and also from deficiencies in software-level control methods. For example, Chinese invention patent application CN109460043A discloses... A self-disturbance rejection control method for ship trajectory based on multimodal non-singular terminal sliding mode is proposed. This method introduces a linear and nonlinear switching extended state observer to estimate the total internal and external disturbances. By looking beyond the surface of the control logic, it is found that the underlying mechanism implicitly relies on the one-dimensional time series error of the lag as the trigger mechanism for state switching and bandwidth adjustment. When facing sudden steep swells with three-dimensional topological asymmetry in real dynamic waters, the simple reliance on the time dimension state deviation feedback mode will inevitably encounter the objective physical mismatch between environmental spatial geometric changes and system time response lag. When waves impact the hull, the algorithm cannot extract the wave geometric deformation characteristics in advance, resulting in a serious lag in the adjustment of control parameters compared with the changes in the physical field gradient, inducing transient yaw and saturation oscillation of the drive mechanism. Instead of seeking compromise within the lag error domain, the method directly extracts the three-dimensional wave characteristics and calculates the spatial curvature, transforming environmental perception from passive lag to spatial forward prediction, thus solving the above-mentioned fundamental spatiotemporal mismatch within its own technical framework.

[0004] Therefore, how to dynamically determine the adjustment parameters based on the spatial geometric distribution characteristics of the water surface disturbance field, and establish a feedback correction mechanism to handle estimation errors under physically constrained working conditions, is the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background section, the technical solution of this invention is as follows:

[0006] An ADRC-based method for navigation planning and control of autonomous sightseeing boats includes the following steps:

[0007] Step S1: Obtain the attitude feedback quantity consisting of the heading angle, speed and position coordinates of the sightseeing boat as the controlled object, as well as the disturbance distribution data representing the water wave morphology within a preset distance ahead of the sightseeing boat, which is collected by the detection load at the front of the sightseeing boat and input through the data interface.

[0008] Step S2: Map the perturbation distribution data to three-dimensional feature sampling points in the perturbation reconstruction domain, and use the three-dimensional feature sampling points as boundary constraints to perform non-uniform rational spline interpolation to fit and generate a perturbation spatiotemporal evolution envelope that continuously represents the gradient of perturbation energy distribution in the future spatiotemporal domain.

[0009] Step S3: Calculate the differential geometric parameters of the perturbation spatiotemporal evolution envelope at the preset prediction point. By extracting the maximum principal curvature of the perturbation spatiotemporal evolution envelope in the current track projection direction, determine the gradient gain factor that characterizes the intensity of environmental perturbation abrupt change and the perturbation spatial deformation characteristics.

[0010] Step S4: Use the gradient gain factor to set the observation bandwidth parameter of the variable bandwidth extended state observer through the bandwidth configuration function, and use the variable bandwidth extended state observer to estimate the lumped disturbance amount including environmental disturbance, unmodeled dynamics inside the sightseeing boat model and nonlinear characteristics of the actuator under the state reconstruction logic.

[0011] Step S5: Based on the real-time deviation between the preset planned path and the pose feedback quantity, an initial adjustment quantity representing the intensity of the control action is generated using a nonlinear deviation feedback control law. The initial adjustment quantity is then processed by feedforward offset compensation using lumped disturbance quantity to output the thrust control command for driving the sightseeing boat's power system.

[0012] Preferably, in step S2, non-uniform rational B-spline logic is used to perform surface modeling of the three-dimensional feature sampling points to generate a perturbation spatiotemporal evolution envelope that continuously represents the distribution of perturbation energy in the future spatiotemporal domain; in step S4, the gradient gain factor is embedded as a feedforward adjustment term into the internal state equation of the variable bandwidth expanded state observer to construct a composite compensation mechanism that coordinates look-ahead prediction and feedback observation.

[0013] Preferably, after step S5, the method further includes: step S6, monitoring the physical limit state of the sightseeing boat's power system at the thrust control command output stage, and calculating the deviation compensation amount between the ideal control quantity and the actual execution quantity; step S7, feeding back the deviation compensation amount to the error correction term of the variable bandwidth expansion state observer, correcting the internal estimation state of the variable bandwidth expansion state observer, and suppressing the observation logic divergence caused by the saturation of the power system.

[0014] Preferably, in step S4, the observation bandwidth parameter The setting rules are as follows: ,in, The preset baseline observation bandwidth is λ, where λ is the sensitivity coefficient. The curvature characteristic value is used; as the curvature characteristic value increases, the variable bandwidth expansion state observer actively increases the observation bandwidth parameter.

[0015] Preferably, the pose feedback includes the real-time position, heading angle, and speed of the sightseeing boat; in step S5, the path tracking deviation is mapped into a heading command using the line-of-sight guidance logic, and the path deviation adjustment logic and dynamic disturbance suppression logic are deeply decoupled to achieve trajectory fidelity control of the sightseeing boat under nonlinear hydrodynamic impact.

[0016] Preferably, in step S1, the disturbance distribution data is acquired by a remote detection payload carried by the sightseeing boat; the acquisition steps include: extracting the three-dimensional coordinates of the wave crests and troughs of the wave motion within a preset distance range ahead, encapsulating the three-dimensional coordinates into a feature matrix characterizing the spatial disturbance energy distribution, and establishing the prior input of the variable bandwidth expansion state observer to the external environmental disturbance.

[0017] Preferably, in step S3, the normal curvature distribution of the perturbation spatiotemporal evolution envelope in the current track projection direction is calculated, and the local maxima in the normal curvature distribution are selected as curvature characteristic values ​​to characterize the steepness of the waves that the sightseeing boat will encounter, providing a geometric basis for bandwidth adjustment for the variable bandwidth expansion state observer.

[0018] Preferably, the nonlinear deviation feedback control law includes a tracking differentiator; in step S5, the tracking differentiator is used to perform smooth transition processing on the planned path, and the first-order rate of change information of the heading command is extracted to generate an advance control gain that eliminates the lag deviation of the sightseeing boat's dynamic response, thereby improving the tracking accuracy of the adjustment system for changes in path curvature.

[0019] Preferably, the thrust control command is used to drive the dual-side pod propulsion system of the sightseeing boat; step S5 further includes: according to the thrust control command, setting the differential speed ratio and heading deflection angle of the dual-side pod propulsion system through the power distribution rules, and using the combined adjustment effect of differential thrust and deflection torque to maintain the dynamic steady-state navigation of the sightseeing boat.

[0020] Preferably, in step S7, when the thrust control command reaches the physical limit threshold of the sightseeing boat's power system, the feedback gain of the nonlinear deviation feedback control law is reduced proportionally according to the deviation compensation amount. The real-time adjustment of the feedback gain is used to limit the integral accumulation effect of the control loop and prevent the sightseeing boat from generating heading sway caused by the accumulation of feedback deviation.

[0021] Compared with existing technologies, the autonomous driving sightseeing boat navigation planning and control method based on ADRC of the present invention has the following advantages:

[0022] 1. In the navigation planning of the autonomous sightseeing boat in ADRC, a parameterized surface is constructed using non-uniform rational B-spline logic. The maximum principal curvature of the disturbance surface is extracted as a topology factor, so that the control parameters and the spatial gradient of the disturbance are dynamically mapped. Existing technologies usually use the root mean square energy of the environmental signal to adjust the control bandwidth. This processing method simplifies the physical disturbance with three-dimensional spatial evolution into a one-dimensional time series scalar, losing the spatial geometric information of the disturbance field. Since the maximum principal curvature accurately reflects the transient gradient of the physical disturbance, when the system faces sudden surges with equivalent energy but steep shape, it can actively converge the observation bandwidth according to the spatial geometric characteristics, suppress high-frequency energy penetration into the control loop, and reduce the transient overshoot of the control system in dynamic waters.

[0023] 2. The collected forward wave state feature tensor is embedded as a feedforward term into the internal state equation of the extended state observer to construct a collaborative compensation mechanism based on spatial look-ahead information and feedback observation information. This design overcomes the inherent time lag defect of inertial sensor data, enabling the system to change its perception of environmental disturbances from a delayed response to a look-ahead prediction. The spatial distribution data provided by the wave feature tensor participates in state reconstruction at the logic layer, enabling the system to generate matching cancellation commands instantly when impacted by waves, eliminating the initial deviation caused by disturbances and improving the path tracking fidelity of sightseeing boats in complex sea conditions.

[0024] 3. Anti-integral saturation back-calculation logic is embedded in the control command output stage. The truncation deviation between the ideal command and the actual driving command is fed back to the error correction term of the observer, establishing a feedback closed loop between the algorithm logic and the hardware physical limit. This mechanism solves the problem of observer logic divergence caused by physical dead zone or rate saturation of the actuator. When a drastic change in disturbance causes driving limit, the real-time injection of the truncation deviation corrects the internal estimated state of the observer, ensuring that the lumped disturbance estimate output by the observer matches the physical reality, avoiding peak oscillations in the observer state, and maintaining the closed-loop stability of the system under extreme conditions. Attached Figure Description

[0025] Figure 1 This is a diagram showing the main steps of the anti-disturbance navigation planning for the autonomous sightseeing boat of this invention;

[0026] Figure 2 This is a logic diagram of the module interaction and closed-loop compensation of the sightseeing boat navigation control system of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0028] An ADRC-based method for navigation planning and control of autonomous sightseeing boats includes the following steps:

[0029] Step S1: Obtain the attitude feedback quantity consisting of the heading angle, speed and position coordinates of the sightseeing boat as the controlled object, as well as the disturbance distribution data representing the water wave morphology within a preset distance ahead of the sightseeing boat, which is collected by the detection load at the front of the sightseeing boat and input through the data interface.

[0030] Step S2: Map the perturbation distribution data to three-dimensional feature sampling points in the perturbation reconstruction domain, and use the three-dimensional feature sampling points as boundary constraints to perform non-uniform rational spline interpolation to fit and generate a perturbation spatiotemporal evolution envelope that continuously represents the gradient of perturbation energy distribution in the future spatiotemporal domain.

[0031] Step S3: Calculate the differential geometric parameters of the perturbation spatiotemporal evolution envelope at the preset prediction point. By extracting the maximum principal curvature of the perturbation spatiotemporal evolution envelope in the current track projection direction, determine the gradient gain factor that characterizes the intensity of environmental perturbation abrupt change and the perturbation spatial deformation characteristics.

[0032] Step S4: Use the gradient gain factor to set the observation bandwidth parameter of the variable bandwidth extended state observer through the bandwidth configuration function, and use the variable bandwidth extended state observer to estimate the lumped disturbance amount including environmental disturbance, unmodeled dynamics inside the sightseeing boat model and nonlinear characteristics of the actuator under the state reconstruction logic.

[0033] Step S5: Based on the real-time deviation between the preset planned path and the pose feedback quantity, an initial adjustment quantity representing the intensity of the control action is generated using a nonlinear deviation feedback control law. The initial adjustment quantity is then processed by feedforward offset compensation using lumped disturbance quantity to output the thrust control command for driving the sightseeing boat's power system.

[0034] Preferably, in step S2, non-uniform rational B-spline logic is used to perform surface modeling of the three-dimensional feature sampling points to generate a perturbation spatiotemporal evolution envelope that continuously represents the distribution of perturbation energy in the future spatiotemporal domain; in step S4, the gradient gain factor is embedded as a feedforward adjustment term into the internal state equation of the variable bandwidth expanded state observer to construct a composite compensation mechanism that coordinates look-ahead prediction and feedback observation.

[0035] Preferably, after step S5, the method further includes: step S6, monitoring the physical limit state of the sightseeing boat's power system at the thrust control command output stage, and calculating the deviation compensation amount between the ideal control quantity and the actual execution quantity; step S7, feeding back the deviation compensation amount to the error correction term of the variable bandwidth expansion state observer, correcting the internal estimation state of the variable bandwidth expansion state observer, and suppressing the observation logic divergence caused by the saturation of the power system.

[0036] Preferably, in step S4, the observation bandwidth parameter The setting rules are as follows: ,in, The preset baseline observation bandwidth is λ, where λ is the sensitivity coefficient. The curvature characteristic value is used; as the curvature characteristic value increases, the variable bandwidth expansion state observer actively increases the observation bandwidth parameter.

[0037] Preferably, the pose feedback includes the real-time position, heading angle, and speed of the sightseeing boat; in step S5, the path tracking deviation is mapped into a heading command using the line-of-sight guidance logic, and the path deviation adjustment logic and dynamic disturbance suppression logic are deeply decoupled to achieve trajectory fidelity control of the sightseeing boat under nonlinear hydrodynamic impact.

[0038] Preferably, in step S1, the disturbance distribution data is acquired by a remote detection payload carried by the sightseeing boat; the acquisition steps include: extracting the three-dimensional coordinates of the wave crests and troughs of the wave motion within a preset distance range ahead, encapsulating the three-dimensional coordinates into a feature matrix characterizing the spatial disturbance energy distribution, and establishing the prior input of the variable bandwidth expansion state observer to the external environmental disturbance.

[0039] Preferably, in step S3, the normal curvature distribution of the perturbation spatiotemporal evolution envelope in the current track projection direction is calculated, and the local maxima in the normal curvature distribution are selected as curvature characteristic values ​​to characterize the steepness of the waves that the sightseeing boat will encounter, providing a geometric basis for bandwidth adjustment for the variable bandwidth expansion state observer.

[0040] Preferably, the nonlinear deviation feedback control law includes a tracking differentiator; in step S5, the tracking differentiator is used to perform smooth transition processing on the planned path, and the first-order rate of change information of the heading command is extracted to generate an advance control gain that eliminates the lag deviation of the sightseeing boat's dynamic response, thereby improving the tracking accuracy of the adjustment system for changes in path curvature.

[0041] Preferably, the thrust control command is used to drive the dual-side pod propulsion system of the sightseeing boat; step S5 further includes: according to the thrust control command, setting the differential speed ratio and heading deflection angle of the dual-side pod propulsion system through the power distribution rules, and using the combined adjustment effect of differential thrust and deflection torque to maintain the dynamic steady-state navigation of the sightseeing boat.

[0042] Preferably, in step S7, when the thrust control command reaches the physical limit threshold of the sightseeing boat's power system, the feedback gain of the nonlinear deviation feedback control law is reduced proportionally according to the deviation compensation amount. The real-time adjustment of the feedback gain is used to limit the integral accumulation effect of the control loop and prevent the sightseeing boat from generating heading sway caused by the accumulation of feedback deviation.

[0043] Example 1: In a dynamic waterway navigation scenario disturbed by strong crosswinds and irregular overlapping waves, a sightseeing boat faces the task of tracking its course and speed trajectory. Conventional autonomous driving posture adjustment systems extract the root mean square energy of the time series of external disturbance signals as the trigger threshold for adjusting the observer bandwidth. This data dimensionality reduction mechanism loses information about the spatial topological asymmetry of the physical wave field. As a result, when the system encounters swells with equivalent energy but steeper morphology, it cannot resolve environmental structural deformation at the data layer, leading to a mismatch in control bandwidth configuration. This causes high-frequency noise penetration of the observer and transient yaw phenomena where the power system exceeds physical limits. An autonomous driving sightseeing boat navigation planning and control method based on ADRC obtains the sightseeing boat's trajectory as the controlled object. The ship's heading angle, speed, and position coordinates constitute the pose feedback quantity, and the disturbance distribution data, characterized by the wave morphology of the water area within a preset distance ahead of the sightseeing ship, is collected by the forward detection payload of the sightseeing ship and input through the data interface. The disturbance distribution data is mapped to three-dimensional feature sampling points in the disturbance reconstruction domain. The three-dimensional feature sampling points are used as boundary constraints and substituted into a non-uniform rational spline interpolation operator to fit and generate a disturbance spatiotemporal evolution envelope that continuously represents the gradient of the disturbance energy distribution in the future spatiotemporal domain. The normal curvature distribution of the disturbance spatiotemporal evolution envelope at the preset prediction point is calculated, and the local maxima in the normal curvature distribution are extracted as curvature feature values. The curvature feature values ​​are used as gradient gain factors for quantifying the spatial disturbance gradient.

[0044] The observation bandwidth parameter of the variable bandwidth extended state observer is set according to the gradient gain factor. The calculation formula is as follows: ,in, This indicates the real-time observed bandwidth parameter. λ represents the preset baseline observation bandwidth, and λ represents the sensitivity coefficient. The curvature characteristic value is represented by the variable bandwidth expansion state observer. As the curvature characteristic value increases, the observation bandwidth parameter is increased. Under the state reconstruction logic, the lumped disturbance quantity, including environmental interference, unmodeled dynamics within the sightseeing boat model, and nonlinear characteristics of the actuator, is estimated. Based on the real-time deviation between the preset planned path and the pose feedback quantity, the first-order rate of change information of the heading command is extracted using the nonlinear deviation feedback control law to generate the initial adjustment quantity characterizing the intensity of the control action. The lumped disturbance quantity is used to apply feedforward offset compensation to the initial adjustment quantity, and the thrust control command driving the sightseeing boat power system is output. Based on the principle of planar kinematic projection, the autopilot system extracts the position coordinates from the pose feedback quantity, calculates the perpendicular orthogonal projection distance from the coordinates to the planned path, and sets it as the lateral tracking error. Based on the geometric guidance formula Calculate the expected heading angle ,in The local tangent inclination angle of the planned path at the corresponding projection point is represented by Δ, which represents the constant forward sight distance. The thrust control command is issued at the node, and the thrust control command is analyzed into the total longitudinal thrust acting on the center of mass of the hull based on the principle of rigid body dynamics torque balance. With head rocking moment For the physical constraints of the dual-pod propulsion system, based on the formula and Deconstruct the underlying execution parameters, among which, and These represent the independent thrust outputs to the left and right pods, respectively. B represents the lateral physical distance constant between the propulsion axes of the left and right pods. The pre-set propeller external characteristic mapping table is extracted to convert the single-sided thrust output into a low-level hardware-matched motor speed control signal. The physical limit state of the sightseeing boat's power system is monitored at the thrust control command output stage. When environmental swells cause the thrust control command to reach the physical limit threshold of the power system, the deviation compensation between the ideal control quantity and the actual execution quantity is calculated and fed back to the error correction term of the variable bandwidth expansion state observer to correct the internal estimation state of the variable bandwidth expansion state observer. This command limit dynamic back-calculation mechanism uses the actual physical boundary error to cut off the observation logic divergence chain caused by the saturation of the actuator rate. Based on the convergence of the observation bandwidth according to the spatial geometric characteristics, the integral accumulation effect of the adjustment loop is limited, so that the sightseeing boat maintains differential thrust and yaw torque output when encountering nonlinear hydrodynamic impacts.

[0045] Example 2: This example establishes a co-simulation verification platform comprising a computational fluid dynamics simulation model and a 6-DOF ship dynamics solver. This platform solves the fundamental Navier-Stokes governing equations to generate a nonlinear wave physical field. JONSWAP wave spectrum signals are input to the platform's boundary conditions, and Gaussian white noise with a signal-to-noise ratio of 20 dB is simultaneously superimposed. This Gaussian white noise is used to simulate measurement errors and physical disturbances of the front-end probe load. The computational parameters of the variable bandwidth extended state observer are set. The basic observation bandwidth parameter is controlled by the natural frequency of the ship's rigid body motion and the response delay of the propulsion system. When the hydrodynamic response delay index increases, the basic observation bandwidth parameter is lowered to suppress high-frequency flutter. The sensitivity coefficient is determined based on the maximum allowable thrust gradient change rate of the propulsion system. For a 15m long sightseeing ship physical model, the basic observation bandwidth parameters are set... The value is 5.0 rad / s, and the sensitivity coefficient λ is set to 1.5 m.

[0046] Extract wave physical field distribution data containing measurement errors, and calculate curvature eigenvalues ​​with three increasing gradients. 0.15m respectively -1 0.52m -1 and 1.18m-1 These correspond to gentle swell conditions, moderate wave conditions, and extremely steep swell conditions, respectively. A control group using a constant observation bandwidth parameter and an experimental group using a variable bandwidth extended state observer were set up. The observation bandwidth parameter of the control group was maintained at 5.0 rad / s, and the experimental group followed the formula... Solve for real-time observation bandwidth parameters ,in, This indicates the real-time observed bandwidth parameter. λ represents the basic observation bandwidth parameter, and λ represents the sensitivity coefficient. Representing the curvature eigenvalues, the experimental group calculated and output the real-time observation bandwidth parameters for the curvature eigenvalues ​​of the three gradients mentioned above. The real-time observation bandwidth parameters are 6.12 rad / s, 8.90 rad / s, and 13.85 rad / s, respectively, and these parameters are applied to the reconstruction logic of the lumped disturbance.

[0047] The root mean square error of heading tracking and the thrust variance are extracted as evaluation parameters, with a curvature characteristic value of 0.15m. -1 Under gentle swell conditions, the root mean square error of heading tracking for the control group and the test group were 2.1° and 1.8°, respectively, and the thrust variance was 12.5 kN. 2 With 13.2kN 2 With a curvature characteristic value of 0.52m -1 Under moderate wave conditions, the root mean square error of the heading tracking in the control group became 4.5°, and the thrust variance increased to 45.8 kN. 2 The root mean square error of the heading tracking in the test group was 2.2°, and the thrust variance was 18.4 kN. 2 With a curvature characteristic value of 1.18m -1 Under extreme steep swell conditions, the thrust control command reaches the physical limit threshold set by the power system, resulting in nonlinear truncation, and the thrust variance of the control group increases to 158.6 kN. 2 The root mean square error of the heading tracking was 9.8°. The test group calculated the deviation compensation between the ideal control quantity and the actual execution quantity and fed it back to the internal estimate state. Its thrust variance was 45.3 kN. 2 The root mean square error of the heading tracking is 3.5°. The constant observation bandwidth parameter leads to the saturation of the power system rate and the integral deviation of the control loop state under the limit disturbance gradient. The test group extracts the wave space curvature feature value as the feedforward adjustment parameter, proportionally amplifies the observation frequency band in the geometrically steep wave impact area, and limits the accumulation of integral error in the saturation area by physical amplitude limiting feedback deviation compensation, thereby weakening the penetration effect of high frequency physical disturbance and outputting convergent differential thrust and heading deflection control commands.

[0048] Example 3: In a dynamic waterway navigation scenario disturbed by irregular waves, the autonomous driving system needs to convert discrete wave detection coordinates into continuous curvature feature factors. Conventional surface fitting operators assign equal fitting weights to each discrete coordinate. When steep swells cause abrupt changes in the density of the detection point cloud in a localized area, the surface reconstruction logic induces local normal vector distortion, resulting in abnormally high-frequency components in the extracted principal curvature values. This causes oscillations in the observation bandwidth parameter of the variable bandwidth expansion state observer. The sightseeing boat's autonomous driving system acquires wave disturbance data output from the front-end detection payload, which includes three-dimensional spatial coordinates and corresponding reflection intensities. The autonomous driving system sets the three-dimensional spatial coordinates as a sequence of control vertices on a non-uniform rational B-spline surface and calculates the wave energy density based on the reflection intensities. The image represents the weighting factor for each control vertex. For control vertices in steep swell regions where the reflection intensity exceeds a set reflection threshold, the autonomous driving system sets a weighting factor greater than the base weight. The system combines node vectors and basis function recursion operators to fit a continuous perturbation spatiotemporal evolution envelope, calculating the first and second partial derivatives of this envelope with respect to the spatial grid parameters. Based on the first partial derivatives, the system solves for the first fundamental form coefficient matrix of the perturbation spatiotemporal evolution envelope, and then solves for the second fundamental form coefficient matrix using the outer product of the first and second partial derivatives. Finally, the system calculates the eigenvalues ​​of the principal curvature equation using the first and second fundamental form coefficient matrices, selecting the largest real root as the curvature eigenvalue at the corresponding prediction point. .

[0049] The autonomous driving system will use curvature characteristic values Substitute the observation bandwidth parameter update logic into the output real-time observation bandwidth parameter. The variable bandwidth expansion state observer is based on this real-time observed bandwidth parameter. The system reconstructs the lumped disturbance quantity including external wave disturbances. The autonomous driving system extracts the heading deviation and its first-order time derivative between the preset planned path and the pose feedback quantity. This heading deviation and first-order time derivative are then input into a nonlinear deviation feedback control law. The nonlinear deviation feedback control law uses a continuous power function to generate an initial adjustment quantity characterizing the control action strength. The autonomous driving system sets the nonlinear gain inflection point of the continuous power function based on the physical response dead zone threshold of the sightseeing boat's steering mechanism. When the heading deviation is within the physical response dead zone threshold, the continuous power function outputs a linear gain characteristic; when the heading deviation exceeds the physical response dead zone threshold... When the value is reached, the continuous power function outputs nonlinear decay gain characteristics. The autonomous driving system uses the lumped disturbance to apply feedforward offset compensation to the initial adjustment amount and outputs thrust control commands. Based on the non-uniform rational B-spline surface reconstruction mechanism that allocates control vertex mapping weights according to reflection intensity, the asymmetric distribution information of wave field spatial energy is preserved. Based on the analytical solution logic of the first and second basic form coefficient matrices, the curvature characteristic value is output. This data processing logic eliminates geometric parameter distortion caused by abrupt changes in the density of discrete data point clouds and outputs differential thrust control commands under the constraint of a nonlinear deviation feedback control law with physical dead zone constraints.

[0050] Example 4: In the engineering calibration scenario of the initial deployment of an autonomous sightseeing boat, the autonomous driving system sends a stepped-incremental open-loop test signal to the dual-pod propulsion system in the initial state where the tolerance of the propulsion mechanism is unknown, and simultaneously collects the hull's yaw rate response data; extracts the critical moment when the yaw rate response data deviates from the zero-point baseline and exceeds the preset noise tolerance, and records the corresponding signal amplitude as the physical response dead zone threshold; sends the ultimate step thrust command, records the mechanical saturation limit when the output shaft reaches the maximum steady-state thrust, sets it as the physical amplitude limit threshold, and establishes the initial physical boundary mapping between the underlying hardware and the control algorithm.

[0051] The autonomous driving system initiates the baseline parameter calibration program of the variable bandwidth extended state observer based on the initial physical boundary mapping, injecting a continuously varying simulated wave disturbance sequence into the control loop; it monitors the convergence curve of the estimation error in the low-frequency band, and extracts the critical gain value that causes the estimation error to converge to a steady state and maintains the non-resonant state of the mechanical structure as the basic observation bandwidth parameter. As the simulated wave disturbance sequence enters the high-frequency band, the autopilot system extracts the reference curvature feature value based on the built-in standard wave model. Under the set condition that the thrust output is lower than the physical amplitude limit threshold, it calculates the dimensionless bandwidth adjustment ratio when the estimation error is at its minimum value. The sensitivity coefficient λ is calculated by using the quotient of the dimensionless bandwidth adjustment ratio and the reference curvature feature value. Combined with the measured physical boundary features, the hardware tolerance is compensated, and the observer baseline parameters adapted to the specific hull are output.

[0052] Example 5: When an autonomous sightseeing boat is deployed in a multipath reflection interference water scene and faces the condition of missing wave reflection intensity perception baseline, the autonomous driving system extracts the spatial point cloud within the calm water time window, calculates the diffuse reflectance variance, establishes the clutter mean within the variance stable interval as the base background noise, drives the hull to sail towards a shock source with known steepness, collects the peak reflection characteristics under the impact of the swell, and uses the difference distribution gradient between it and the base background noise to construct a signal-to-noise ratio matrix. The autonomous driving system extracts the intensity parameters corresponding to the attenuation boundary points based on the signal-to-noise ratio matrix as the set reflection threshold for dividing the control vertex weights, and outputs the signal to the quantization scale of the wave energy density image.

[0053] The autonomous driving system extracts the navigation boundary and divides it into an initial 2D mesh. For sparse point cloud regions at the edge of the sensor's field of view, virtual control vertices are generated based on the neighborhood energy conservation mechanism. Compensation interpolation is applied to the edge nodes of the initial 2D mesh along the trajectory projection direction to generate a closed node sequence. The autonomous driving system injects a penalty factor into the partial derivative solution matrix of the first and second fundamental forms of the surface to suppress boundary curvature divergence. When sampling is missing, the curvature eigenvalues ​​are affected. When the error exceeds the preset tolerance range, the autonomous driving system freezes the observation bandwidth parameter at the effective convergence value, compensates for the physical blind spot, and outputs the perturbation spatiotemporal evolution envelope that maintains the geometric continuity constraint.

[0054] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A navigation planning and control method for an autonomous sightseeing boat based on ADRC, characterized in that, Includes the following steps: Step S1: Obtain the attitude feedback quantity consisting of the heading angle, speed and position coordinates of the sightseeing boat as the controlled object, as well as the disturbance distribution data representing the water wave morphology within a preset distance ahead of the sightseeing boat, which is collected by the detection load at the front of the sightseeing boat and input through the data interface. Step S2: Map the perturbation distribution data to three-dimensional feature sampling points in the perturbation reconstruction domain, and use the three-dimensional feature sampling points as boundary constraints to perform non-uniform rational spline interpolation to fit and generate a perturbation spatiotemporal evolution envelope that continuously represents the gradient of perturbation energy distribution in the future spatiotemporal domain. Step S3: Calculate the differential geometric parameters of the perturbation spatiotemporal evolution envelope at the preset prediction point. By extracting the maximum principal curvature of the perturbation spatiotemporal evolution envelope in the current track projection direction, determine the gradient gain factor that characterizes the intensity of environmental perturbation abrupt change and the perturbation spatial deformation characteristics. Step S4: Use the gradient gain factor to set the observation bandwidth parameter of the variable bandwidth extended state observer through the bandwidth configuration function, and use the variable bandwidth extended state observer to estimate the lumped disturbance amount including environmental disturbance, unmodeled dynamics inside the sightseeing boat model and nonlinear characteristics of the actuator under the state reconstruction logic. Step S5: Based on the real-time deviation between the preset planned path and the pose feedback quantity, an initial adjustment quantity representing the intensity of the control action is generated using a nonlinear deviation feedback control law. The initial adjustment quantity is then fed forward to compensate for the lumped disturbance quantity, so as to output the thrust control command for driving the sightseeing boat's power system. In step S2, non-uniform rational B-spline logic is used to perform surface modeling of the three-dimensional feature sampling points to generate a perturbation spatiotemporal evolution envelope that continuously represents the distribution of perturbation energy in the future spatiotemporal domain; in step S4, the gradient gain factor is embedded as a feedforward adjustment term into the internal state equation of the variable bandwidth extended state observer to construct a composite compensation mechanism that coordinates look-ahead prediction and feedback observation. In step S3, the normal curvature distribution of the perturbation spatiotemporal evolution envelope in the current track projection direction is calculated, and the local maxima in the normal curvature distribution are selected as curvature characteristic values ​​to characterize the steepness of the waves that the sightseeing boat will encounter, providing a geometric basis for bandwidth adjustment for the variable bandwidth expansion state observer. In step S4, the observation bandwidth parameter is... The setting rules are as follows: ,in, The preset baseline observation bandwidth is λ, where λ is the sensitivity coefficient. The curvature characteristic value is used; as the curvature characteristic value increases, the variable bandwidth expansion state observer actively increases the observation bandwidth parameter. Step S5 is followed by: Step S6, monitoring the physical limit state of the sightseeing boat's power system at the thrust control command output stage, and calculating the deviation compensation amount between the ideal control quantity and the actual execution quantity; Step S7, feeding back the deviation compensation amount to the error correction term of the variable bandwidth expansion state observer, correcting the internal estimation state of the variable bandwidth expansion state observer, and suppressing the observation logic divergence caused by the saturation of the power system.

2. The method for navigation planning and control of an autonomous sightseeing boat based on ADRC as described in claim 1, characterized in that, The pose feedback includes the sightseeing boat's real-time position, heading angle, and speed. In step S5, the path tracking deviation is mapped to a heading command using the line-of-sight guidance logic, and the path deviation adjustment logic and dynamic disturbance suppression logic are deeply decoupled to achieve trajectory fidelity control of the sightseeing boat under nonlinear hydrodynamic impact.

3. The method for navigation planning and control of an autonomous sightseeing boat based on ADRC as described in claim 1, characterized in that, In step S1, the disturbance distribution data is acquired through the remote detection payload carried on the sightseeing boat; The acquisition steps include: extracting the three-dimensional coordinates of the wave crests and troughs within a preset distance range ahead, encapsulating the three-dimensional coordinates into a feature matrix characterizing the energy distribution of spatial disturbances, and establishing a priori input for the variable bandwidth expansion state observer to external environmental disturbances.

4. The method for navigation planning and control of an autonomous sightseeing boat based on ADRC as described in claim 1, characterized in that, The nonlinear deviation feedback control law includes a tracking differentiator; in step S5, the tracking differentiator is used to smooth the planned path and extract the first-order rate of change information of the heading command to generate an advance control gain that eliminates the lag deviation of the sightseeing boat's dynamic response, thereby improving the tracking accuracy of the regulation system to changes in path curvature.

5. The method for navigation planning and control of an autonomous sightseeing boat based on ADRC as described in claim 1, characterized in that, The thrust control command is used to drive the dual-side pod propulsion system of the sightseeing boat; step S5 also includes: according to the thrust control command, setting the differential speed ratio and heading deflection angle of the dual-side pod propulsion system through the power distribution rules, and using the combined adjustment effect of differential thrust and deflection torque to maintain the dynamic steady-state navigation of the sightseeing boat.

6. The method for navigation planning and control of an autonomous sightseeing boat based on ADRC as described in claim 1, characterized in that, In step S7, when the thrust control command reaches the physical limit threshold of the sightseeing boat's power system, the feedback gain of the nonlinear deviation feedback control law is reduced proportionally according to the deviation compensation amount. The real-time adjustment of this feedback gain is used to limit the integral accumulation effect of the control loop and prevent the sightseeing boat from generating heading sway caused by the accumulation of feedback deviation.

Images