Filtering experience planning rotating drum sailboat double-layer trigger control method, device and equipment

Abstract

The application discloses a filter experience planning rotating drum sailboat double-layer trigger control method, device and equipment, relates to the field of ship motion control research. The main method comprises the following steps: a three-degree-of-freedom actual ship motion model and a virtual ship kinematics mathematical model of a rotating drum sailboat are constructed, and a reference path of the virtual ship is obtained; a path tracking error is determined, a virtual control law is generated based on the path tracking error; the virtual control law is filtered based on a dynamic surface control, and a dynamic error is generated; an actuator gain is estimated online by an adaptive law based on an updated evaluation network and the dynamic error to generate a control strategy; the control strategy is triggered and evaluated based on a double-layer dynamic integral event trigger condition containing a state cumulative factor, and a control instruction is updated and output to an actuator. The application can realize comprehensive control of high-precision path tracking, online learning stability and actuator protection under complex marine environment.

Description

Technical Field

[0001] This invention relates to the field of ship motion control research, specifically to a filtered empirical planning dual-layer trigger control method, device, and equipment for a rotary sailboat. Background Technology

[0002] With the increasing demand for energy conservation and emission reduction in the shipping industry, rotor-assisted vessels (RAVs) have attracted widespread attention due to their high energy efficiency. RAVs rely on the Magnus effect, generating a pressure difference during rotor rotation to produce auxiliary wind thrust, making it a highly efficient and energy-saving propulsion solution. Existing research has significantly improved aerodynamic performance through optimizing the rotor layout. However, during ship navigation, there is a strong coupling between the rotor, hull, and ocean disturbances. Therefore, the path tracking control problem of RAVs under random ocean disturbances still faces the following two challenges: First, for the path tracking control problem of underactuated unmanned vessels, traditional empirical playback-based adaptive dynamic programming algorithms typically require persistent excitation (PE) as a prerequisite for network parameter convergence. This persistent excitation condition essentially demands that the system's state and control signals possess sufficient spectral richness throughout the entire operation to support the evaluation network's gradual convergence to the global optimum in fitting the value function. This rigid constraint leads to unnecessary oscillations in the system state and control inputs, increasing not only the wasted energy consumption and mechanical wear of the ship's propulsion system but also directly causing a decrease in system robustness and trajectory tracking accuracy.

[0003] Secondly, most existing event triggering conditions employ a single-layer triggering mechanism based on instantaneous states, relying solely on the comparison between the state deviation at the current sampling moment and a threshold to determine whether to update the control law. This type of triggering logic struggles to effectively distinguish between steady-state deviations and instantaneous deviations caused by short-term ocean disturbances. Therefore, such mechanisms commonly suffer from problems like false triggering and trigger oscillations, exhibiting low resistance to instantaneous ocean disturbances. Ultimately, this leads to reduced control accuracy during long-term ship navigation, and very limited improvement in actuator wear. Summary of the Invention

[0004] To address the problems of stringent online learning convergence conditions leading to system oscillations and frequent actuator wear caused by sudden threshold changes in event triggering mechanisms in existing technologies, this application proposes a filtered empirical planning dual-layer triggering control method, device, and equipment for rotary sailboats. This achieves comprehensive control that balances high-precision path tracking, online learning stability, and actuator protection in complex marine environments.

[0005] The technical means employed in this invention are as follows: A filtered empirical planning method for dual-layer trigger control of a rotary sailboat includes the following steps: S1. Construct a three-degree-of-freedom real-ship motion model of a rotary sailboat, which includes a rotary sail auxiliary thrust term, an unknown nonlinear dynamics term, and an ocean time-varying disturbance term. S2. Set waypoint information according to the navigation mission, construct a virtual ship kinematic mathematical model based on the virtual ship guidance principle of L1, and obtain the reference path of the virtual ship based on the virtual ship kinematic mathematical model. S3. Obtain ship state data based on the three-degree-of-freedom real ship motion model of the rotary sailboat, and determine the path tracking error based on the ship state data and the reference path; the path tracking error includes position error and heading angle error, and generate a virtual control law based on the path tracking error; S4. The virtual control law is filtered based on dynamic surface control to generate a dynamic surface signal, and the dynamic error is calculated based on the dynamic surface signal. S5. The system uncertainty is approximated by an evaluation network, the weights of which are updated based on a filter-cumulative empirical replay mechanism; the actuator gain is estimated online using an adaptive law based on the updated evaluation network and the dynamic error to generate an initial control strategy. S6. Based on the two-layer dynamic integral event triggering condition including the state accumulation factor, determine whether the initial control strategy needs to be updated. When the triggering condition is met, update and output the control command to the actuator to control the ship to perform path tracking.

[0006] Furthermore, the three-degree-of-freedom real ship motion model is shown in formula (1): (1) in, This indicates the planar position of the rotary sailboat in the inertial coordinate system. This represents the bow angle of the rotary sailboat in an inertial coordinate system. This represents the forward speed, drift speed, and bow roll rate of the rotary sailboat in the attached coordinate system. , and Indicates the positive inertial mass coefficient. , and This represents the model uncertainty term for the rotary sailboat in its forward, sideways, and bow-rolling degrees of freedom. , and This indicates the propeller thrust in response to marine environmental disturbances in the forward, drift, and yaw directions. satisfy ,in Given the propeller speed, the yaw torque is defined as follows: ,in For rudder angle, coefficient and This represents the gain function of the system actuator. This indicates the auxiliary thrust of the rotary sail; The auxiliary thrust of the rotary sail Calculate according to formula (2): (2) in, Indicates air density, and These represent the apparent wind speed and apparent wind direction, respectively. This represents the maximum projected area of ​​the rotary sail. These represent the lift and drag coefficients of the rotary sail, respectively.

[0007] Furthermore, the mathematical model of the virtual ship's kinematics is shown in formula (3): (3) in, and These represent the virtual ship's position and heading angle, respectively. For the virtual ship's forward speed, This represents the bow roll rate of the virtual ship; The reference path of the virtual ship is obtained based on the aforementioned kinematic mathematical model of the virtual ship, including: Based on waypoint information, a reference path is generated by smoothly connecting adjacent waypoints using straight line segments and circular arc segments; both straight line segments and circular arc segments are determined based on the turning angle between adjacent waypoints. The azimuth angle of the virtual ship relative to the real ship is used as the guidance rate, as shown in formula (4): (4) In the formula, The virtual ship's position coordinates, These are the actual coordinates of the ship's position. It is a symbolic function; Depending on the type of interval to which the reference path belongs (i.e., a straight segment or a circular segment), the perpendicular distance between the virtual ship and the waypoint is used. As a criterion, when When positive, it generates a velocity perpendicular to the forward speed. Lateral velocity: For the straight segment, the lateral velocity is determined according to formula (5): (5) In the formula, For the virtual ship's forward speed, This represents the forward distance between the virtual ship's current position and a reference point on the desired trajectory. Represents the virtual ship's velocity vector and The angle between the connecting lines, where, This represents the expected bow angle of the virtual ship. for and The angle between the axes, where Represents the horizontal axis in a nautical coordinate system; For the circular arc segment, the lateral velocity is determined according to formula (6): (6) in, Indicates the first The virtual ship's forward speed in the arc segment of each waypoint Indicates the first The arc segment at each waypoint represents the virtual ship's forward speed. and The angle between the lines, The radius of the arc is denoted as .

[0008] Furthermore, a virtual control law is generated based on the path tracking error, including: Based on formula (7), construct a virtual control law for the design position error and heading error:

[0009] (7) in, For positional error, For heading angle error and For positive design parameters, For overshoot constant, Indicates guidance rate.

[0010] Furthermore, the virtual control law is filtered based on dynamic surface control to generate a dynamic surface signal, and the dynamic error is calculated based on the dynamic surface signal, including: Dynamic surface control is performed according to formula (8):

[0011] (8) in, The filter time constant is greater than zero. Represents dynamic surface signals in the forward and yaw degrees of freedom; Define the dynamic error in the forward and yaw degrees of freedom as: and .

[0012] Furthermore, the system uncertainty is approximated by an evaluation network, the weights of which are updated based on a filtered cumulative empirical replay mechanism; based on the updated evaluation network and the dynamic error, the actuator gain is estimated online using an adaptive law to generate an initial control strategy, including: The dynamic errors in the forward and yaw degrees of freedom are differentiated according to formula (9):

[0013] (9) In the formula, and The uncertainty term represents the overall expression, and the online approximation of the comment network is performed using formula (10): (10) In the formula, These represent the comment network weights on the forward and pitching degrees of freedom, respectively. This represents the corresponding Gaussian function. , Indicates the approximation error. These represent the estimated comment network weights for the forward and yaw degrees of freedom, respectively. Construct adaptive functions for forward and yaw degrees of freedom. and The estimated value and To compensate for the inherent uncertainty in the actuator gain, the actuator control law is given according to formula (11), and the corresponding adaptive law is given according to formula (12):

[0014] (11)

[0015] (12) In the formula, Positive design parameters For the combined thrust, it is represented as , This represents the partial derivative of the Gaussian function. This indicates the propeller speed.

[0016] Furthermore, the weight update law of the comment network is given according to formula (13):

[0017] (13) in, For positive design parameters, Satisfying equation (14),

[0018] (14) In the formula, , , , The filtered signal is represented by equation (15):

[0019]

[0020] (15) In the formula, , , , .

[0021] Furthermore, based on a two-layer dynamic integral event triggering condition including a state accumulation factor, the initial control strategy is triggered and evaluated. In response to the satisfaction of the triggering condition, control commands are updated and output to the actuators to control the vessel to perform path tracking, including: The state accumulation factor is introduced to trigger the two-layer dynamic integral event according to equations (16)-(18):

[0022] (16) Triggering condition is

[0023] (17) State cumulative factor , and dynamic integral term , Represented as

[0024]

[0025]

[0026] (18) In the formula, Indicates positive design parameters. For system output error and , and Indicates input error. and For a moment and The propeller speed and rudder angle at that time and For a moment and The propeller speed and rudder angle at that time.

[0027] This invention also discloses a filtered empirical planning rotary sailboat dual-layer trigger control device, used to implement the method described in any one of the above, comprising: The motion model construction unit is used to construct a three-degree-of-freedom real-ship motion model of a rotary sailboat, which includes rotary sail auxiliary thrust, unknown nonlinear dynamics and ocean time-varying disturbances. The reference path acquisition unit is used to set waypoint information according to the navigation mission, construct a virtual ship kinematic mathematical model based on the L1 virtual ship guidance principle, and acquire the reference path of the virtual ship based on the virtual ship kinematic mathematical model. The virtual control law generation unit is used to acquire ship state data based on the three-degree-of-freedom real ship motion model of the rotary sailboat, determine the path tracking error based on the ship state data and the reference path; the path tracking error includes position error and heading angle error, and generate a virtual control law based on the path tracking error; The dynamic error acquisition unit is used to filter the virtual control law based on dynamic surface control, generate a dynamic surface signal, and calculate and acquire the dynamic error based on the dynamic surface signal. A control strategy generation unit is used to approximate system uncertainties through an evaluation network, the weights of which are updated based on a filtered cumulative empirical replay mechanism; and to generate a control strategy by online estimation of actuator gain using an adaptive law based on the updated evaluation network and the dynamic error. The path tracking unit is used to trigger and evaluate the control strategy based on a two-layer dynamic integral event triggering condition that includes a state accumulation factor. In response to the satisfaction of the triggering condition, it updates and outputs control commands to the actuator to control the ship to perform path tracking.

[0028] The present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the filtering empirical planning rotary sailboat dual-layer trigger control method described in any of the above claims.

[0029] Compared with the prior art, the present invention has the following advantages: 1. This invention proposes a filtered empirical programming two-layer trigger control method for RAV path tracking, specifically for rotary sailboats. The method first obtains a reference path by constructing a three-degree-of-freedom real-ship motion model and a virtual ship kinematic mathematical model based on the L1 virtual ship guidance principle. This determines the path tracking error, including position and heading angle errors, and generates a virtual control law. Subsequently, the virtual control law is filtered using dynamic surface control to generate a dynamic error. An evaluation network approximates the system uncertainty, where the weights are updated based on a filtered cumulative empirical playback mechanism, effectively converting instantaneous signals into cumulative signals and alleviating the stringent continuous excitation conditions in online learning. Simultaneously, based on the updated evaluation network and the dynamic error, an adaptive law is used to estimate the actuator gain online to generate an initial control strategy. Finally, the initial control strategy is triggered and evaluated based on a two-layer dynamic integral event trigger condition including a state accumulation factor. When the trigger condition is met, the control command is updated and output. Therefore, this invention can balance high-precision path tracking, online learning stability, and actuator protection in complex marine environments, significantly improving the stability and robustness of ship operation.

[0030] 2. This invention proposes a dual-layer dynamic integral event triggering mechanism to achieve smooth triggering of control commands, effectively solve the mechanical fatigue problem caused by frequent actuator actions, enhance the anti-interference capability of the system, and ultimately achieve a 16.56% reduction in energy consumption, making it more suitable for engineering application needs in actual marine navigation scenarios. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart of the ship path tracking control signal in an embodiment of the present invention.

[0033] Figure 2 This is an example of the Magnus effect and force analysis of a rotary sailboat in an embodiment of the present invention, where (a) indicates that the auxiliary thrust of the rotary sail originates from the Magnus effect, and (b) shows a schematic diagram of the forces acting on the rotary sailboat.

[0034] Figure 3 This is a schematic diagram of the guidance principle of the L1 virtual ship in an embodiment of the present invention.

[0035] Figure 4This is a conceptual signal flow diagram of the dual-layer trigger control method for a rotary sailboat based on dynamic programming with filtering experience in an embodiment of the present invention.

[0036] Figure 5 This is the tracking trajectory of a rotary sailboat under a random wind field in an embodiment of the present invention.

[0037] Figure 6 This is the path tracking error curve of the dual-layer trigger control method for a rotary sailboat based on dynamic programming with filtering experience in an embodiment of the present invention.

[0038] Figure 7 This is the control input curve of the dual-layer trigger control method for a rotary sailboat based on dynamic programming with filtering experience in this embodiment of the invention.

[0039] Figure 8 This is a comparative analysis of the thrust performance and energy optimization ratio of the rotary sail in an embodiment of the present invention.

[0040] Figure 9 This is a schematic diagram of the rotational speed of the rotary sail in an embodiment of the present invention.

[0041] Figure 10 This is a comparison of the path tracking trajectories of the method of the present invention and the comparative method in the embodiments of the present invention.

[0042] Figure 11 This is a comparison of the positional and heading errors between the method of the present invention and a comparative method in this embodiment of the invention.

[0043] Figure 12 This is a comparison of the propeller speed and rudder angle control inputs between the method of the present invention and a comparative method in this embodiment of the invention.

[0044] Figure 13 This is a comparison of propeller speed and rudder angle trigger interval between the method of the present invention and a comparative method in this embodiment of the invention. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0046] Example 1 like Figures 1-4As shown, this invention provides a two-layer triggered control method for a rotary sailboat using filtered empirical planning. Combining the L1 virtual ship guidance principle, it constructs a two-layer dynamic integral event-triggered control method based on filtered cumulative empirical playback adaptive dynamic programming to achieve energy-saving path tracking of the RAV under random ocean disturbances. Specifically, the method of this invention mainly includes the following steps: S1. Construct a three-degree-of-freedom real-ship motion model of a rotary sailboat, which includes the auxiliary thrust of the rotary sail, unknown nonlinear dynamics, and ocean time-varying disturbances.

[0047] Specifically, this embodiment is based on the classical underactuated surface ship dynamics framework and constructs a three-degree-of-freedom motion model that incorporates the auxiliary thrust of the swivel sail, unknown nonlinear dynamics and time-varying ocean disturbances, as shown in Equation (19).

[0048] (19) In the formula, the state vector Indicates the planar position of RAV in the inertial coordinate system and heading angle . This represents the forward velocity, drift velocity, and yaw rate of the RAV in the attached coordinate system. , and This represents the positive inertial mass coefficient. , and This represents the model uncertainty term for the rotary sailboat in terms of forward, sideways, and bow-rolling degrees of freedom. , and This indicates disturbance to the marine environment. The propeller thrust satisfies... ,in Given the propeller speed, the yaw torque is defined as follows: ,in For rudder angle. Coefficient. and This represents the gain function of the system actuator. This indicates the auxiliary thrust of the rotary sail.

[0049] As attached Figure 2 As shown in (a), the auxiliary thrust of the rotary sail originates from the Magnus effect. Under crosswind conditions, a fluid pressure difference is generated around the rotary sail. This generates a net aerodynamic force perpendicular to the windward direction, providing auxiliary forward thrust for the ship.

[0050] As attached Figure 2 As shown in (b), the figure illustrates the forces acting on the rotary sailboat. For nautical coordinate systems, This represents the ship's appendage coordinate system, where L and D represent the lift and drag forces acting on the rotating drum, respectively. and These are the auxiliary thrust and lift acting on the ship, respectively. and These represent the true wind speed and the true wind direction, respectively. Let be the ship's forward speed. Based on this, the auxiliary thrust generated by the magnus effect of the swivel can be expressed as: (20) In the formula, This indicates the auxiliary thrust of the rotary sail. Indicates air density, and These represent the apparent wind speed and apparent wind direction, respectively. This indicates the maximum projected area of ​​the rotary sail. These represent the lift and drag coefficients of the rotary sail, respectively.

[0051] S2. Set waypoint information according to the navigation mission, construct a virtual ship kinematic mathematical model based on the virtual ship guidance principle of L1, and obtain the reference path of the virtual ship based on the virtual ship kinematic mathematical model.

[0052] Specifically, in order to generate a continuous and smooth reference path, this embodiment adopts the L1-based virtual ship guidance principle, the schematic diagram of which is attached. Figure 3 As shown, where, For nautical coordinate systems, Indicates the ship's accessory coordinate system. , , Let m-1, m, and m+1 be the waypoints, and u and v be the ship's forward speed and lateral drift speed, respectively. Indicates the actual ship's heading angle. For actual ship heading angle and guidance rate The difference is given below, along with the kinematic mathematical model of the L1 virtual ship.

[0053] (twenty one) In the formula, and These represent the virtual ship's position and heading angle, respectively. This represents the forward speed of the virtual ship. This represents the bow roll rate of the virtual ship.

[0054] For a real ship, the guidance rate is defined as the azimuth angle of the virtual ship relative to the real ship, as shown in equation (22).

[0055] (twenty two) In the formula, The virtual ship's position coordinates, These are the actual coordinates of the ship's position. It is a symbolic function.

[0056] The reference path is divided into straight segments and circular segments. To avoid overshoot issues on straight segments, the perpendicular distance between the virtual ship and the waypoint is used. As a criterion, this mechanism ensures that the virtual ship's reference path always converges to the waypoint path. Specifically, when When positive, it will generate a velocity perpendicular to the forward speed. Lateral speed command As shown in equation (23).

[0057] (twenty three) In the formula, This represents the forward distance between the virtual ship's current position and a reference point on the desired trajectory. Represents the virtual ship's velocity vector and The angle between the connecting lines, where, This represents the expected bow angle of the virtual ship. for and The angle between axes.

[0058] For the circular arc segment, the lateral velocity is determined by formula (24).

[0059] (twenty four) According to geometric properties, Substituting this into the above equation yields the virtual ship's turning speed. . This represents the forward speed of the virtual ship. Indicates the first The virtual ship's forward speed is represented by the arc segment of each waypoint. Indicates the first The arc segment at each waypoint represents the virtual ship's forward speed. and The angle between the lines, Indicates the virtual ship in the 1st month The expected heading angle of each waypoint For the first waypoints and The angle between axes. The radius of the arc is denoted as .

[0060] S3. Obtain ship state data based on the three-degree-of-freedom real ship motion model of the rotary sailboat, and determine the path tracking error based on the ship state data and the reference path; the path tracking error includes position error and heading angle error, and generate a virtual control law based on the path tracking error.

[0061] Specifically, in terms of controller design, this embodiment proposes an adaptive dynamic programming algorithm based on filtered experience playback, and a dynamic integral event triggering mechanism including a state accumulation factor. By combining the backstepping method with optimal control theory, a path tracking algorithm that utilizes a single evaluation network to achieve adaptive online updates is proposed. Specifically, firstly, in order to stabilize the ship's position error... and heading angle error Design virtual control law and .

[0062]

[0063] (25) In the formula, and These are positive design parameters. To keep it small and constant, so as to avoid ship overshoot.

[0064] S4. The virtual control law is filtered based on dynamic surface control to generate a dynamic surface signal, and the dynamic error is calculated based on the dynamic surface signal.

[0065] Specifically, to avoid the problem of excessive computational complexity in the traditional backstepping method, Dynamic Surface Control (DSC) technology is introduced, as shown in equation (26).

[0066]

[0067] (26) In the formula, The filter time constant is greater than zero. This represents the dynamic surface signal in the forward and yaw degrees of freedom. The difference between the dynamic surface signal and the virtual control law can be expressed as: ,Right now .

[0068] Further define the dynamic errors in the forward and yaw degrees of freedom as: and , S5. The system uncertainty is approximated by an evaluation network, the weights of which are updated based on a filtered cumulative empirical replay mechanism; the actuator gain is estimated online using an adaptive law based on the updated evaluation network and the dynamic error to generate an initial control strategy.

[0069] Specifically, differentiating the dynamic errors in the forward and yaw degrees of freedom yields the following:

[0070] (27) In the formula, and The uncertain terms representing the synthesis can be approximated online through a comment network, as shown in equation (9).

[0071] (28) In the formula, These represent the comment network weights on the forward and pitching degrees of freedom, respectively. This represents the corresponding Gaussian function. , Indicates the approximation error. These represent the estimated comment network weights for the forward and pitching degrees of freedom, respectively.

[0072] Construct adaptive functions for forward and yaw degrees of freedom. and The estimated value and Used to compensate for inherent uncertainties in actuator gain. Actuator control law. , , , As given by equation (29), the corresponding adaptive law is equation (30).

[0073]

[0074] (29)

[0075] (30) In the formula, The design parameters are positive. For the combined thrust, it is represented as , This represents the partial derivative of the Gaussian function.

[0076] The weight update law of the comment network can be designed according to the gradient descent method as Equation (31).

[0077]

[0078] (31) in, For positive design parameters, It satisfies equation (32).

[0079]

[0080] (32) In the formula, , , . The filtered signal is represented by equation (33).

[0081]

[0082]

[0083] (33) In the formula, , , , , which are the corresponding accumulated experience values.

[0084] S6. Based on the two-layer dynamic integral event triggering condition including the state accumulation factor, the initial control strategy is triggered and evaluated. In response to the satisfaction of the triggering condition, the control command is updated and output to the actuator to control the ship to perform path tracking.

[0085] Specifically, this embodiment proposes a two-layer dynamic integral event triggering mechanism, introducing a state accumulation factor to address high-frequency noise and instantaneous ocean disturbances, thereby further reducing ineffective actuator actions and extending equipment lifespan. The specific triggering mechanism is shown in equations (34)-(36).

[0086]

[0087] (34) Triggering condition is

[0088] (35) State cumulative factor , and dynamic integral term , Represented as

[0089]

[0090]

[0091] (36) In the formula, Indicates positive design parameters. For system output error and , and Indicates input error. and For a moment and The propeller speed and rudder angle at that time and For a moment and The propeller speed and rudder angle are measured at the specified time. Finally, the final speed and rudder angle can be obtained using a MATLAB servo system. , .

[0092] To verify the beneficial effects of this invention, two sets of representative experiments were conducted on the MATLAB simulation platform to comprehensively verify the control performance of the proposed algorithm: 1) The performance and energy efficiency of the proposed algorithm were evaluated under random ocean disturbance conditions; 2) A comparative study was conducted with the tracking control algorithm in the literature "Shen Z, Zhou L, Yang Y, et al. Dynamic Event-Triggered ADPOptimized Tracking Control of Marine Surface Vessels With Asymmetric Time-Varying Full-State Constraint[J]. IEEE Transactions on Intelligent Transportation Systems, 2024, 25(7): 7352-7362."

[0093] The trajectory of the RAV is as follows Figure 5 As shown, the marine environment corresponding to sea state 4 was simulated using the standard NORSOK wind spectrum and JONSWAP wave spectrum, and additional disturbances were introduced through a stochastic wind field model. Under these combined environmental disturbances, the RAV sequentially passed through all preset waypoints while maintaining smooth route changes and high-precision path tracking. Figure 6It shows the error in the ship's position and heading angle. Figure 7 This paper demonstrates the actuator input under a two-layer dynamic integral event-triggered control scheme. This mechanism eliminates threshold abrupt changes and enhances the system's robustness to ocean disturbances. Technical improvements make control input variations more stable and effectively prevent excessive actuator wear. Figure 8 The comparison curves of propeller thrust and rotary sail auxiliary thrust are shown, along with the Energy Optimization Rate (EOR). Calculations show that using the rotary sail assist technology reduces energy consumption by approximately 16.56% over the entire voyage. Figure 9 The adaptive speed adjustment process of the rotary sail in the wind field is demonstrated.

[0094] Figures 10-13 The experimental results comparing the algorithm proposed in this invention with the prior art in the literature "Shen Z, Zhou L, Yang Y, et al. Dynamic Event-Triggered ADP Optimized Tracking Control of Marine SurfaceVessels With Asymmetric Time-Varying Full-State Constraint[J]. IEEE Transactions on Intelligent Transportation Systems, 2024, 25(7): 7352-7362." are presented. Figure 10 The trajectories of the two algorithms are shown. Both algorithms achieve good tracking performance, but the proposed algorithm outperforms them in terms of tracking accuracy and dynamic response speed. Figure 11 The curves showing the position and heading angle errors under the two algorithms are presented. The proposed algorithm exhibits smoother error variation. Figure 12 The control input curves of the two algorithms are shown. Compared with the event-triggered mechanism of the comparison method, the proposed algorithm has a lower oscillation frequency and fewer executor command updates. Figure 13 The difference in trigger interval between the two algorithms is presented, and the proposed algorithm has significant advantages.

[0095] Example 2 This embodiment discloses a filtered empirical planning rotary sailboat dual-layer trigger control device, which is a functional modular encapsulation of the filtered empirical planning rotary sailboat dual-layer trigger control method in Embodiment 1. Specifically, the device includes the following functional modules.

[0096] The motion model construction unit is used to construct a three-degree-of-freedom real-ship motion model of a rotary sailboat, which includes rotary sail auxiliary thrust, unknown nonlinear dynamics and ocean time-varying disturbances. The reference path acquisition unit is used to set waypoint information according to the navigation mission, construct a virtual ship kinematic mathematical model based on the L1 virtual ship guidance principle, and acquire the reference path of the virtual ship based on the virtual ship kinematic mathematical model. The virtual control law generation unit is used to acquire ship state data based on the three-degree-of-freedom real ship motion model of the rotary sailboat, determine the path tracking error based on the ship state data and the reference path; the path tracking error includes position error and heading angle error, and generate a virtual control law based on the path tracking error; The dynamic error acquisition unit is used to filter the virtual control law based on dynamic surface control, generate a dynamic surface signal, and calculate and acquire the dynamic error based on the dynamic surface signal. A control strategy generation unit is used to approximate system uncertainties through an evaluation network, the weights of which are updated based on a filtered cumulative empirical replay mechanism; and to generate a control strategy by online estimation of actuator gain using an adaptive law based on the updated evaluation network and the dynamic error. The path tracking unit is used to trigger and evaluate the control strategy based on a two-layer dynamic integral event triggering condition that includes a state accumulation factor. In response to the satisfaction of the triggering condition, it updates and outputs control commands to the actuator to control the ship to perform path tracking.

[0097] Example 3 This embodiment discloses a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the filtering empirical planning rotary sailboat dual-layer trigger control method described in Embodiment 1.

[0098] Specifically, this embodiment maps the abstracted method flows and functional modules of the foregoing embodiments onto specific physical hardware carriers, thereby providing a physical implementation foundation for the technical solution. The electronic device here does not refer to a specific model or brand of controller, but rather encompasses all general-purpose or special-purpose computing devices with data storage and instruction execution capabilities.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A filtered empirical planning dual-layer trigger control method for a rotary sailboat, characterized in that, Includes the following steps: S1. Construct a three-degree-of-freedom real-ship motion model of a rotary sailboat, which includes a rotary sail auxiliary thrust term, an unknown nonlinear dynamics term, and an ocean time-varying disturbance term. S2. Set waypoint information according to the navigation mission, construct a virtual ship kinematic mathematical model based on the virtual ship guidance principle of L1, and obtain the reference path of the virtual ship based on the virtual ship kinematic mathematical model. S3. Obtain ship state data based on the three-degree-of-freedom real ship motion model of the rotary sailboat, and determine the path tracking error based on the ship state data and the reference path; the path tracking error includes position error and heading angle error, and generate a virtual control law based on the path tracking error; S4. The virtual control law is filtered based on dynamic surface control to generate a dynamic surface signal, and the dynamic error is calculated based on the dynamic surface signal. S5. The system uncertainty is approximated by an evaluation network, the weights of which are updated based on a filter-cumulative experience replay mechanism; Based on the updated evaluation network and the dynamic error, the actuator gain is estimated online using an adaptive law to generate an initial control strategy; S6. Based on the two-layer dynamic integral event triggering condition including the state accumulation factor, determine whether the initial control strategy needs to be updated. When the triggering condition is met, update and output the control command to the actuator to control the ship to perform path tracking.

2. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 1, characterized in that, The three-degree-of-freedom ship motion model is shown in formula (1): (1) in, This indicates the planar position of the rotary sailboat in the inertial coordinate system. This represents the bow angle of the rotary sailboat in an inertial coordinate system. This represents the forward speed, drift speed, and bow roll rate of the rotary sailboat in the attached coordinate system. , and Indicates the positive inertial mass coefficient. , and This represents the model uncertainty term for the rotary sailboat in its forward, sideways, and bow-rolling degrees of freedom. , and This indicates the propeller thrust in response to marine environmental disturbances in the forward, drift, and yaw directions. satisfy ,in Given the propeller speed, the yaw torque is defined as follows: ,in For rudder angle, coefficient and This represents the gain function of the system actuator. This indicates the auxiliary thrust of the rotary sail; The auxiliary thrust of the rotary sail Calculate according to formula (2): (2) in, Indicates air density, and These represent the apparent wind speed and apparent wind direction, respectively. This represents the maximum projected area of ​​the rotary sail. These represent the lift and drag coefficients of the rotary sail, respectively.

3. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 2, characterized in that, The mathematical model of the virtual ship's kinematics is shown in formula (3): (3) in, and These represent the virtual ship's position and heading angle, respectively. For the virtual ship's forward speed, This represents the bow roll rate of the virtual ship; The reference path of the virtual ship is obtained based on the aforementioned kinematic mathematical model of the virtual ship, including: Based on waypoint information, a reference path is generated by smoothly connecting adjacent waypoints using straight line segments and circular arc segments; both straight line segments and circular arc segments are determined based on the turning angle between adjacent waypoints. The azimuth angle of the virtual ship relative to the real ship is used as the guidance rate, as shown in formula (4): (4) In the formula, The virtual ship's position coordinates, These are the actual coordinates of the ship's position. It is a symbolic function; Depending on the type of interval to which the reference path belongs (i.e., a straight segment or a circular segment), the perpendicular distance between the virtual ship and the waypoint is used. As a criterion, when When positive, it generates a velocity perpendicular to the forward speed. Lateral velocity: For the straight segment, the lateral velocity is determined according to formula (5): （5） In the formula, For the virtual ship's forward speed, This represents the forward distance between the virtual ship's current position and a reference point on the desired trajectory. Represents the virtual ship's velocity vector and The angle between the connecting lines, where, This represents the expected bow angle of the virtual ship. for and The angle between the axes, where Represents the horizontal axis in a nautical coordinate system; For the circular arc segment, the lateral velocity is determined according to formula (6): （6） in, Indicates the first The virtual ship's forward speed in the arc segment of each waypoint Indicates the first The arc segment at each waypoint represents the virtual ship's forward speed. and The angle between the lines, The radius of the arc is denoted as .

4. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 3, characterized in that, A virtual control law is generated based on the path tracking error, including: Based on formula (7), a virtual control law based on position error and heading error is constructed: (7) in, For positional error, For heading angle error and For positive design parameters, For overshoot constant, Indicates guidance rate.

5. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 4, characterized in that, The virtual control law is filtered based on dynamic surface control to generate a dynamic surface signal, and the dynamic error is calculated based on the dynamic surface signal, including: Dynamic surface control is performed according to formula (8): (8) in, The filter time constant is greater than zero. Represents dynamic surface signals in the forward and yaw degrees of freedom; Define the dynamic error in the forward and yaw degrees of freedom as: and .

6. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 5, characterized in that, The evaluation network approximates the system uncertainty, and the weights of the evaluation network are updated based on a filter-cumulative empirical replay mechanism. Based on the updated evaluation network and the dynamic error, the actuator gain is estimated online using an adaptive law to generate an initial control strategy, including: The dynamic errors in the forward and yaw degrees of freedom are differentiated according to formula (9): (9) In the formula, and The uncertainty term represents the overall expression, and the online approximation of the comment network is performed using formula (10): (10) In the formula, These represent the comment network weights on the forward and pitching degrees of freedom, respectively. This represents the corresponding Gaussian function. , Indicates the approximation error. These represent the estimated comment network weights for the forward and yaw degrees of freedom, respectively. Construct adaptive functions for forward and yaw degrees of freedom. and The estimated value and To compensate for the inherent uncertainty in the actuator gain, the actuator control law is given according to formula (11), and the corresponding adaptive law is given according to formula (12): (11) (12） In the formula, Positive design parameters For the combined thrust, it is represented as , This represents the partial derivative of the Gaussian function. This indicates the propeller speed.

7. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 6, characterized in that, The weight update law of the comment network is given according to formula (13): (13) in, For positive design parameters, Satisfying equation (14), (14) In the formula, , , , The filtered signal is represented by equation (15): (15) In the formula, , , , .

8. The filtered empirical planning dual-layer trigger control method for a rotary sailboat according to claim 7, characterized in that, Based on a two-layer dynamic integral event triggering condition that includes a state accumulation factor, the initial control strategy is triggered and evaluated. In response to the satisfaction of the triggering condition, control commands are updated and output to the actuators to control the vessel to perform path tracking, including: The state accumulation factor is introduced to trigger the two-layer dynamic integral event according to equations (16)-(18): (16) Triggering condition is (17) State cumulative factor , and dynamic integral term , Represented as (18) In the formula, Indicates positive design parameters. For system output error and , and Indicates input error. and For a moment and The propeller speed and rudder angle at that time and For a moment and The propeller speed and rudder angle at that time.

9. A filter-based empirical planning rotary sailboat dual-layer trigger control device, used to implement the method as described in any one of claims 1-8, characterized in that, include: The motion model construction unit is used to construct a three-degree-of-freedom real-ship motion model of a rotary sailboat, which includes rotary sail auxiliary thrust, unknown nonlinear dynamics and ocean time-varying disturbances. The reference path acquisition unit is used to set waypoint information according to the navigation mission, construct a virtual ship kinematic mathematical model based on the L1 virtual ship guidance principle, and acquire the reference path of the virtual ship based on the virtual ship kinematic mathematical model. The virtual control law generation unit is used to acquire ship state data based on the three-degree-of-freedom real ship motion model of the rotary sailboat, determine the path tracking error based on the ship state data and the reference path; the path tracking error includes position error and heading angle error, and generate a virtual control law based on the path tracking error; The dynamic error acquisition unit is used to filter the virtual control law based on dynamic surface control, generate a dynamic surface signal, and calculate and acquire the dynamic error based on the dynamic surface signal. A control strategy generation unit is used to approximate system uncertainties through an evaluation network whose weights are updated based on a filter-cumulative experience replay mechanism. Based on the updated evaluation network and the dynamic error, the actuator gain is estimated online using an adaptive law to generate a control strategy; The path tracking unit is used to trigger and evaluate the control strategy based on a two-layer dynamic integral event triggering condition that includes a state accumulation factor. In response to the satisfaction of the triggering condition, it updates and outputs control commands to the actuator to control the ship to perform path tracking.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the filtering empirical planning rotary sailboat dual-layer trigger control method according to any one of claims 1 to 8.

Images