Unmanned ship berthing path tracking control system and method in narrow water area of port
By combining convolutional neural networks and nonlinear model predictive control, the accuracy and stability issues of berthing control for unmanned vessels in narrow port waters were solved. This resulted in high-precision, smooth, and safe berthing path tracking, reduced energy consumption and mechanical wear, and improved the control performance of unmanned vessels in complex environments.
Patent Information
- Application Number
- CN202511610841.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-05
- Publication Date
- 2025-12-26
AI Technical Summary
Existing technologies are insufficient to meet the requirements of high precision, stability and safety in the berthing control of unmanned vessels in narrow waters of ports. Especially in low-speed, shallow water and multi-disturbance environments, traditional methods are difficult to accurately characterize the nonlinear characteristics of ships, resulting in ship deviation or attitude instability, as well as high energy consumption and mechanical wear.
By combining a convolutional neural network (CNN) module with a nonlinear model predictive control (NMPC) module, the CNN is used to extract features from historical path data to generate reference path points. The nonlinear model predictive control is then used to optimize the solution of the ship's motion state. At the same time, the rolling time domain estimation (MHE) module is used to correct the state estimation, improve robustness, and achieve high-precision and smooth berthing path tracking control.
It enables unmanned vessels to berth with high precision, stability and safety in complex port environments, reduces the frequent operation of propellers and servo motors, reduces energy consumption and mechanical wear, and improves port berth scheduling efficiency and operational safety.
Smart Images

Figure CN121209515A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned vessel automated control and intelligent shipping technology, and more particularly to a berthing path tracking control system and method for unmanned vessels in narrow waterways of ports. Background Technology
[0002] With the rapid development of intelligent shipping technology, unmanned vessels (UAVs) are increasingly widely used in port operations. Autonomous berthing of UAVs is a key component in achieving port automation and intelligence. In the narrow waters of ports, UAVs need to perform operations such as berthing, navigation scheduling, and inspection and monitoring. These tasks place extremely high demands on the path tracking control of UAVs. Path tracking control of UAVs involves multiple aspects, including ship motion modeling, path planning, and collision avoidance algorithms, with the aim of ensuring that UAVs can complete berthing operations with high precision, stability, and safety.
[0003] Currently, research in the field of unmanned surface vessels (USVs) mainly focuses on ship motion modeling, path planning, and collision avoidance algorithms. Research on autonomous berthing control in the narrow waters of ports is relatively insufficient. Traditional control methods such as PID (Proportional-Integral-Derivative Control), LOS (Linear Observer System), or MPC (Model Predictive Control) are widely used in USV berthing control. These methods can achieve path tracking and berthing control of USVs to a certain extent, but in the complex port environment of low-speed navigation, shallow water effects, and multi-source disturbances, their performance often fails to meet the requirements for high-precision, smooth, and safe berthing. Furthermore, existing technologies also have limitations in terms of control energy utilization efficiency and robustness. For example, the ship motion model is described using the three-degree-of-freedom linear state-space model proposed by Fossen, which takes the following form:
[0004] In the formula, For the mass and additional mass matrix, The linear damping and restoring force matrix, For the ship's velocity vector, This model is used for control inputs. It is computationally simple and easy to design controllers, but it is difficult to accurately characterize the nonlinear characteristics of ships under low speed, shallow water, and wind and wave disturbances.
[0005] To improve modeling accuracy, the four-degree-of-freedom nonlinear model of the controlled object is obtained from the above equation, as shown in the following equation:
[0006]
[0007] In the formula, The matrix represents the Coriolis force and the centrifugal force. Here is the damping parameter matrix. It is a wind and wave interference matrix, reflecting the impact of wind and waves on longitudinal velocity, lateral velocity, yaw rate, and roll rate. This represents the shallow water effect interference matrix. These are the position and attitude vectors. This model can accurately characterize the nonlinear dynamics of a ship.
[0008] In existing technologies, studies have employed nonlinear model predictive control or combined with the Line-of-Sight (LOS) algorithm for autonomous berthing control of ships. (Ship status) and control input The nonlinear prediction model can be expressed as:
[0009] NMPC solves for the control input sequence through iterative optimization within the future prediction time domain. This allows the ship to berth along a reference path. The LOS algorithm corrects the course based on the geometric relationship between the current position and the reference path, improving path tracking accuracy in narrow berth environments. Its typical optimization objective function is:
[0010] In the formula, As a reference path point, To predict the time domain, To control the time domain, , These are the state and control weighting matrices, respectively. By minimizing the deviation between the predicted state and the reference path point, and simultaneously suppressing the variation in control input, path tracking and energy consumption balance are achieved.
[0011] While existing technologies have made some progress in unmanned surface vessel (USV) berthing control, several significant shortcomings remain. First, traditional path-following algorithms fail to adequately account for shallow water effects and disturbances such as wind, waves, and currents, leading to lateral drift, yaw, or rolling as the vessel approaches the berth, making accurate berthing difficult. Second, existing controllers are mostly based on linear or simplified dynamic models, resulting in lag in steering and unstable speeds at low speeds, causing the trajectory to deviate from the reference path and affecting berthing accuracy. Finally, existing algorithms are prone to generating excessive control outputs in narrow waterways or under conditions of strong winds, waves, or significant shallow water effects, increasing mechanical wear on propellers and servos and reducing energy efficiency. Therefore, the applicability of existing technologies in the complex environments of narrow port waterways is limited, making it difficult to meet the requirements for high-precision, stable, and safe berthing of USVs. Summary of the Invention
[0012] To address the aforementioned technical problems, this invention provides a control system and method for tracking the berthing path of unmanned vessels in narrow waterways of ports. This invention combines a convolutional neural network (CNN) module with a nonlinear model predictive control (NMPC) module. The CNN module references path features and inputs them into the NMPC module to achieve nonlinear optimization of the control input. This system can accurately predict the vessel's motion state under low-speed, shallow water, and complex port disturbance conditions, optimizing course and propulsion control, enabling the unmanned vessel to berth with high precision and stability along a reference path, thereby improving port berth scheduling efficiency and operational safety.
[0013] The technical means employed in this invention are as follows: A berthing path tracking and control system for unmanned vessels in narrow waterways of a port includes: The Convolutional Neural Network (CNN) module is used to extract features from historical path data, generate reference path points, and use them as input to the controller. The nonlinear model predictive control (NMPC) module is used to optimize the motion state of the unmanned vessel based on the reference path points generated by the convolutional neural network (CNN) module, so that the unmanned vessel gradually converges to the reference path. The Rolling Time-Domain Estimation (MHE) module is used to correct the state estimation of the Nonlinear Model Predictive Control (NMPC) module using historical observation data, thereby improving robustness to wind and wave disturbances and measurement errors, and enabling high-precision, smooth and safe berthing path tracking control of unmanned vessels in narrow waters of ports.
[0014] Furthermore, the topology of the convolutional neural network (CNN) module consists of an input layer, a convolutional layer, a BN-ReLU activation layer, a Dropout layer, a fully connected layer, and an output layer, wherein: The convolutional layer uses a one-dimensional convolutional kernel to extract local features of the path in the time series. After nonlinear transformation by the BN-ReLU activation layer, a stable feature representation is obtained. The Dropout layer is used to reduce the risk of overfitting. The fully connected layer integrates features, and the output layer generates dynamic path prediction points.
[0015] Furthermore, the convolution calculation formula for the convolutional layer is as follows:
[0016] in, Indicates the first The feature maps output by the convolutional layers are located at... The value at, where, and It is a spatial location index on the feature map. Indicates the current convolutional layer number; Indicates the activation function; Indicates the first The number of feature maps in a layer, i.e., the number of input channels; and These represent the height and width of the convolution kernel, respectively; Indicates the first Layer Each convolutional kernel is located at... Weight at each location; Indicates the index of the convolution kernel. and Indicates the position index inside the convolution kernel; Indicates the first The first layer feature map Each channel is located in The value at; This represents the bias term, used to adjust the output of the convolutional layer.
[0017] Furthermore, in the nonlinear model predictive control (NMPC) module, the behavior of the unmanned vessel berthing path tracking control system in the narrow waters of the port is predicted at future time steps, and the control input vector is optimized. To enable the state vector of the unmanned vessel berthing path tracking control system in the narrow waters of the port. The formula for tracing the reference path is as follows:
[0018] in, This represents the state update function; This indicates the output function.
[0019] Furthermore, the rolling time-domain estimation (MHE) module uses historical observation data to correct the state estimation of the nonlinear model predictive control (NMPC) module, specifically including: Define the optimization problem as follows:
[0020]
[0021]
[0022] in, Represents the cost function; Indicates time step The system status; Indicates from time step arrive Control input; Indicates possible disturbances or noise; Indicates system state Control input and disturbance The constraints that must be met; Represents a set of constraints; A multiple-targeting approach is employed, treating both state and control variables as augmented state variables simultaneously. This transforms the path tracking problem into a high-dimensional point stability problem. Within each sampling period, the rolling time-domain estimation (MHE) module corrects the prediction error based on the observation sequence. Its objective function is:
[0023]
[0024]
[0025]
[0026] in, Represent the objective function; A sequence representing state variables; Represents a sequence of control inputs; Indicates the length of the time window for the Rolling Time-Domain Estimation (MHE) module; Indicates at time step At that time, the output value estimated by the Rolling Time-Domain Estimation (MHE) module; Indicates at time step The actual output value measured at that time; express and The weighted sum of squares of the differences between them; Indicates at time step At that time, the control input value is estimated through the optimization process; Indicates at time step At that time, the actual control input value applied; express and The weighted sum of squares of the differences between them; Indicates the current time step; It represents the set of all state variables that satisfy the constraints.
[0027] The present invention also provides a method for tracking and controlling the berthing path of an unmanned vessel in narrow waters of a port, based on the aforementioned unmanned vessel berthing path tracking and control system, comprising: S1. Use the convolutional neural network (CNN) module to extract features from historical path data, generate reference path points and use them as input to the controller; S2. The motion state of the unmanned vessel is optimized and solved using the nonlinear model predictive control (NMPC) module, so that the unmanned vessel gradually converges to the reference path. S3. The rolling time-domain estimation (MHE) module is used to correct the state estimation of the nonlinear model predictive control (NMPC) module based on historical observation data, thereby improving the robustness to wind and wave disturbances and measurement errors, and realizing high-precision, stable and safe berthing path tracking control of unmanned vessels in narrow waters of ports.
[0028] Compared with the prior art, the present invention has the following advantages: 1. This invention extracts features from historical path data using a convolutional neural network (CNN) module, enabling the identification of the characteristic relationship between ship motion and environmental disturbances, generating high-quality reference path points as input to the controller. This technique allows unmanned vessels to maintain high-precision path tracking control even in narrow harbor waters and complex dynamic environments with disturbances such as wind and waves, enabling ships to berth with high precision along the reference track and improving berthing accuracy.
[0029] 2. This invention utilizes a nonlinear model predictive control (NMPC) module to optimize the motion state of the unmanned vessel. This technique can accurately predict the vessel's motion state, optimize heading and propulsion control, further improve the accuracy of path tracking, and ensure that the unmanned vessel can accurately berth along the preset path.
[0030] 3. Under strong external interference or sudden path changes, this invention can quickly adjust the ship's speed and course by combining CNN and NMPC, effectively suppressing overshoot and oscillation, ensuring the stability and continuity of the control process, and ensuring the safety and stability of the unmanned ship in berthing path tracking.
[0031] 4. This invention employs Rolling Time-Domain Estimation (MHE) technology, utilizing historical observations to correct state estimation, thereby improving robustness to wind and wave disturbances and measurement errors. This technique further ensures stable control of the unmanned vessel under complex disturbance conditions, maintaining stable berthing path tracking even in complex situations such as strong winds and waves or shallow water effects.
[0032] 5. Existing technologies are prone to generating excessive control output in narrow waters or when encountering strong winds and waves or significant shallow water effects. However, this invention optimizes control input, reduces frequent and large-amplitude movements of the propeller and servo motor, lowers energy consumption and mechanical wear, extends equipment lifespan, and improves the economic efficiency and engineering application value of unmanned vessel berthing operations.
[0033] 6. This invention uses a multiple-targeting method to simultaneously treat state and control variables as augmented state variables, thereby transforming the path tracking problem into a stability problem of high-dimensional points. This further optimizes the control strategy, improves control efficiency, and reduces unnecessary energy consumption and mechanical wear.
[0034] In summary, this invention solves the problems existing in the tracking and control of unmanned vessels' berthing paths in narrow waters of ports through the above-mentioned technical means, and achieves high-precision, stable and safe tracking and control of unmanned vessels' berthing paths, which has significant technical advantages and practical application value. Attached Figure Description
[0035] 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.
[0036] Figure 1 This is a system block diagram of the present invention.
[0037] Figure 2 This is the topology of the Convolutional Neural Network (CNN) module of this invention.
[0038] Figure 3 This is a flowchart of the method of the present invention.
[0039] Figure 4 The image shows an actual unmanned surface vessel provided for an embodiment of the present invention.
[0040] Figure 5 A comparison of path tracking performance before and after the improvement of the controller provided in this embodiment of the invention.
[0041] Figure 6 A comparison of longitudinal velocity changes before and after the algorithm improvement provided in this embodiment of the invention.
[0042] Figure 7 A comparison of lateral velocity changes before and after the algorithm improvement provided in this embodiment of the invention.
[0043] Figure 8 A comparison of the changes in bow roll angular velocity before and after the algorithm improvement provided in this embodiment of the invention.
[0044] Figure 9 Comparison of roll angular velocity changes before and after algorithm improvement provided in embodiments of the present invention.
[0045] Figure 10 A comparison of track error changes before and after algorithm improvements provided in this embodiment of the invention. Detailed Implementation
[0046] 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 of the present invention. 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.
[0047] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, products or devices.
[0048] like Figure 1 As shown, the present invention provides an unmanned vessel berthing path tracking and control system for narrow waterways in ports, comprising: The Convolutional Neural Network (CNN) module is used to extract features from historical path data, generate reference path points, and use them as input to the controller. The nonlinear model predictive control (NMPC) module is used to optimize the motion state of the unmanned vessel based on the reference path points generated by the convolutional neural network (CNN) module, so that the unmanned vessel gradually converges to the reference path. The Rolling Time-Domain Estimation (MHE) module is used to correct the state estimation of the Nonlinear Model Predictive Control (NMPC) module using historical observation data, thereby improving robustness to wind and wave disturbances and measurement errors, and enabling high-precision, smooth and safe berthing path tracking control of unmanned vessels in narrow waters of ports.
[0049] In specific implementation, as a preferred embodiment of the present invention, such as Figure 2 As shown, the topology of the convolutional neural network (CNN) module consists of an input layer, a convolutional layer, a BN-ReLU activation layer, a Dropout layer, a fully connected layer, and an output layer, wherein: The convolutional layer uses a one-dimensional convolutional kernel to extract local features of the path over time. After nonlinear transformation by the BN-ReLU activation layer, a stable feature representation is obtained. The Dropout layer is used to reduce the risk of overfitting, the fully connected layer integrates the features, and the output layer generates dynamic path prediction points. In this embodiment, by training on a large amount of pre-defined path data, this convolutional neural network (CNN) module can quickly generate high-quality path reference points in new environments, providing dynamic input to the controller.
[0050] In a specific implementation, as a preferred embodiment of the present invention, the convolution calculation formula for the convolutional layer is:
[0051] in, Indicates the first The feature maps output by the convolutional layers are located at... The value at, where, and It is a spatial location index on the feature map. Indicates the current convolutional layer number; Indicates the activation function; Indicates the first The number of feature maps in a layer, i.e., the number of input channels; and These represent the height and width of the convolution kernel, respectively; Indicates the first Layer Each convolutional kernel is located at... Weight at each location; Indicates the index of the convolution kernel. and Indicates the position index inside the convolution kernel; Indicates the first The first layer feature map Each channel is located in The value at; This represents the bias term, used to adjust the output of the convolutional layer.
[0052] In a specific implementation, as a preferred embodiment of the present invention, the nonlinear model predictive control (NMPC) module predicts the behavior of the unmanned vessel berthing path tracking control system in the narrow waters of the port at future time steps, and optimizes the control input vector. To enable the state vector of the unmanned vessel berthing path tracking control system in the narrow waters of the port. The formula for tracing the reference path is as follows:
[0053] in, This represents the state update function; This indicates the output function.
[0054] In a preferred embodiment of the present invention, the rolling time-domain estimation (MHE) module uses historical observation data to correct the state estimation of the nonlinear model predictive control (NMPC) module, specifically including: Define the optimization problem as follows:
[0055]
[0056]
[0057] in, Represents the cost function; Indicates time step The system status; Indicates from time step arrive Control input; Indicates possible disturbances or noise; Indicates system state Control input and disturbance The constraints that must be met; Represents a set of constraints; A multiple-targeting approach is employed, treating both state and control variables as augmented state variables simultaneously. This transforms the path tracking problem into a high-dimensional point stability problem. Within each sampling period, the rolling time-domain estimation (MHE) module corrects the prediction error based on the observation sequence. Its objective function is:
[0058]
[0059]
[0060]
[0061] in, Represent the objective function; A sequence representing state variables; Represents a sequence of control inputs; Indicates the length of the time window for the Rolling Time-Domain Estimation (MHE) module; Indicates at time step At that time, the output value estimated by the Rolling Time-Domain Estimation (MHE) module; Indicates at time step The actual output value measured at that time; express and The weighted sum of squares of the differences between them; Indicates at time step At that time, the control input value is estimated through the optimization process; Indicates at time step At that time, the actual control input value applied; express and The weighted sum of squares of the differences between them; Indicates the current time step; It represents the set of all state variables that satisfy the constraints.
[0062] In this embodiment, the Rolling Time-Domain Estimation (MHE) module solves the optimization problem online and corrects the ship's position by combining the azimuth information, thereby compensating for wind and wave interference and sensor noise.
[0063] like Figure 3 As shown, the present invention also provides a method for tracking and controlling the berthing path of an unmanned vessel in a narrow port area, based on the aforementioned unmanned vessel berthing path tracking and control system, comprising: S1. Use the convolutional neural network (CNN) module to extract features from historical path data, generate reference path points and use them as input to the controller; S2. The motion state of the unmanned vessel is optimized and solved using the nonlinear model predictive control (NMPC) module, so that the unmanned vessel gradually converges to the reference path. S3. The rolling time-domain estimation (MHE) module is used to correct the state estimation of the nonlinear model predictive control (NMPC) module based on historical observation data, thereby improving the robustness to wind and wave disturbances and measurement errors, and realizing high-precision, stable and safe berthing path tracking control of unmanned vessels in narrow waters of ports.
[0064] Example This embodiment uses an unmanned surface vessel (such as...) Figure 4 A simulation model for tracking the berthing path of a ship was established for the controlled object. The controller was implemented using the MATLAB / CasADi platform, with a prediction step size of 80 and a sampling period of 1 s.
[0065] First, the path tracking performance of traditional NMPC, NMPC combined with MHE, and the CNN+NMPC+MHE control method proposed in this invention were compared in complex port environments. Figure 5 As shown, Figure 5 (a) This indicates that traditional methods are prone to path oscillations and deviations during berthing, making it difficult to maintain stability; Figure 5 (b) This indicates that the control system of the present invention can quickly converge to the predetermined path, with a smooth trajectory and significantly reduced deviation, thereby effectively avoiding the risk of ship deviation or unstable attitude during berthing and improving operational safety.
[0066] Further comparisons were made of the control effects of the three methods on longitudinal velocity, lateral velocity, yaw rate, roll rate, and track error. For example... Figure 6-10As shown, the control system of this invention can achieve stable control of longitudinal and lateral speeds during the autonomous berthing of unmanned vessels, avoiding overshoot and ensuring continuous and controllable speed changes during berthing. It exhibits small fluctuations in yaw and roll angular velocity control, significantly improving the safety and reliability of the berthing process. Simultaneously, the system reduces frequent and large-amplitude movements of the propellers and servos, lowering energy consumption and mechanical wear, extending equipment lifespan. It can be directly applied to autonomous berthing operations of unmanned vessels in narrow waterways of ports, demonstrating good engineering practicality and promotional value.
[0067] 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 therein. Such 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 tracking and control system for the berthing path of an unmanned vessel in narrow waters of a port, characterized in that, include: The convolutional neural network module is used to extract features from historical path data, generate reference path points, and use them as input to the controller. The nonlinear model prediction control module is used to optimize the motion state of the unmanned vessel based on the reference path points generated by the convolutional neural network module, so that the unmanned vessel gradually converges to the reference path. The rolling time-domain estimation module is used to correct the state estimation of the nonlinear model predictive control module using historical observation data, thereby improving the robustness to wind and wave disturbances and measurement errors, and enabling high-precision, stable and safe berthing path tracking control of unmanned vessels in narrow waters of ports.
2. The unmanned vessel berthing path tracking and control system in a narrow port waterway according to claim 1, characterized in that, The topology of the convolutional neural network module consists of an input layer, a convolutional layer, a BN-ReLU activation layer, a Dropout layer, a fully connected layer, and an output layer, wherein: The convolutional layer uses a one-dimensional convolutional kernel to extract local features of the path in the time series. After nonlinear transformation by the BN-ReLU activation layer, a stable feature representation is obtained. The Dropout layer is used to reduce the risk of overfitting. The fully connected layer integrates features, and the output layer generates dynamic path prediction points.
3. The unmanned vessel berthing path tracking and control system in a narrow port waterway according to claim 2, characterized in that, The convolution calculation formula for the convolutional layer is: in, Indicates the first The feature maps output by the convolutional layers are located at... The value at, where, and It is a spatial location index on the feature map. Indicates the current convolutional layer number; Indicates the activation function; Indicates the first The number of feature maps in a layer, i.e., the number of input channels; and These represent the height and width of the convolution kernel, respectively; Indicates the first Layer Each convolutional kernel is located at... Weight at each location; Indicates the index of the convolution kernel. and Indicates the position index inside the convolution kernel; Indicates the first The first layer feature map Each channel is located in The value at; This represents the bias term, used to adjust the output of the convolutional layer.
4. The unmanned vessel berthing path tracking and control system in a narrow port waterway according to claim 1, characterized in that, In the nonlinear model predictive control module, the behavior of the unmanned vessel berthing path tracking control system in the narrow waters of the port is predicted at future time steps, and the control input vector is optimized. To enable the state vector of the unmanned vessel berthing path tracking control system in the narrow waters of the port. The formula for tracing the reference path is as follows: in, This represents the state update function; This indicates the output function.
5. The unmanned vessel berthing path tracking and control system for narrow waterways in a port as described in claim 1, characterized in that, The rolling time-domain estimation module uses historical observation data to correct the state estimation of the nonlinear model predictive control module, specifically including: Define the optimization problem as follows: in, Represents the cost function; Indicates time step The system status; Indicates from time step arrive Control input; Indicates possible disturbances or noise; Indicates system state Control input and disturbance The constraints that must be met; Represents a set of constraints; A multiple-targeting approach is employed, treating both state and control variables as augmented state variables, thus transforming the path tracking problem into a high-dimensional point stability problem. Within each sampling period, the rolling time-domain estimation module corrects the prediction error based on the observation sequence. Its objective function is: in, Represent the objective function; A sequence representing state variables; Represents a sequence of control inputs; Indicates the length of the time window for the rolling time-domain estimation module; Indicates at time step At that time, the output value estimated by the rolling time-domain estimation module; Indicates at time step The actual output value measured at that time; express and The weighted sum of squares of the differences between them; Indicates at time step At that time, the control input value is estimated through the optimization process; Indicates at time step At that time, the actual control input value applied; express and The weighted sum of squares of the differences between them; Indicates the current time step; It represents the set of all state variables that satisfy the constraints.
6. A method for tracking and controlling the berthing path of an unmanned vessel in a narrow port waterway, based on the berthing path tracking and control system described in any one of claims 1-5, characterized in that, include: S1. Use a convolutional neural network module to extract features from historical path data, generate reference path points and use them as controller input; S2. The motion state of the unmanned vessel is optimized and solved by the nonlinear model prediction control module, so that the unmanned vessel gradually converges to the reference path. S3. The rolling time-domain estimation module is used to correct the state estimation of the nonlinear model predictive control module based on historical observation data, thereby improving the robustness to wind and wave disturbances and measurement errors, and realizing high-precision, stable and safe berthing path tracking control of unmanned vessels in narrow waters of ports.
Citation Information
Patent Citations
Method for estimating angular velocity of spacecraft based on rolling time domain estimation
CN109343550A
State prediction and estimation method for unmanned vehicle
CN112758097A
Control method applied to autonomous berthing of under-actuated double-propeller double-rudder ship
CN113110468A
Ship berthing and leaving control method and device, electronic equipment and storage medium
CN118244759A
Method for training an artificial neural network model to predict the berthing capacity of ships, method for predicting the berthing capacity of ships and non-transitory computer-readable medium
US20250131265A1
