A high-precision trajectory tracking control method for a water surface robot and related equipment

Abstract

The application belongs to the technical field of control, and discloses a high-precision trajectory tracking control method for a water surface robot and related equipment, the method comprising: constructing a dynamic multi-modal environment map containing three-dimensional geometric topological information and water flow field vector information of the water surface; performing global path planning by using an improved quantum particle swarm optimization algorithm embedded with a water flow dynamics equation constraint according to the dynamic multi-modal environment map, and performing smoothing parameterization processing on the global path obtained by planning to generate a reference trajectory; predicting flow field disturbance based on a fluid-structure coupling dynamics model according to the water flow field vector information, estimating system residual disturbance by using an extended state observer, and combining a generated feedforward control amount; adaptively switching a nonlinear model predictive controller and an adaptive sliding mode controller according to real-time water flow field flow velocity and trajectory tracking error, combining the feedforward control amount, and generating a control instruction to drive the water surface robot to track the reference trajectory; and thus high-precision trajectory tracking control is achieved.

Description

Technical Field

[0001] This application relates to the field of control technology, and more specifically, to a high-precision trajectory tracking control method and related equipment for a surface robot. Background Technology

[0002] Surface robots have broad application prospects in fields such as environmental monitoring, water quality sampling, and underwater exploration. However, in practical applications, surface robots often face complex aquatic environments, such as water surface reflections, dynamically floating obstacles, and complex and variable water flow fields. These factors seriously affect the autonomous operation capabilities and trajectory tracking accuracy of surface robots.

[0003] Existing autonomous operation control strategies for surface robots primarily rely on traditional image vision detection algorithms for environmental perception. However, these traditional algorithms are susceptible to specular reflection interference in environments with strong water surface glare, leading to decreased accuracy in detecting floating objects, and even missed or false detections, thus reducing the environmental adaptability of surface robots. Furthermore, floating objects on the water surface are often dynamic, making traditional static obstacle detection methods ineffective, further increasing the difficulty of path planning and obstacle avoidance for surface robots.

[0004] In path planning, traditional geometry-based methods often fail to adequately consider the dynamic coupling effect between fluid and the robot itself. This means that when planning a path, the robot's own motion characteristics and the influence of water flow on the robot are not effectively integrated, resulting in high energy consumption and difficulty in accurate tracking of the planned trajectory during actual execution. Especially in environments with complex water flow fields, this planning method often fails to generate optimal, energy-efficient paths.

[0005] In trajectory tracking control, traditional PID controllers and MPC model predictive controllers often exhibit delayed responses to complex hydrodynamic disturbances such as nonlinear surges and abrupt flow fields, leading to significant trajectory tracking errors. These controllers have limitations in handling strongly nonlinear, time-varying systems and external disturbances, making it difficult to meet the high-precision trajectory tracking requirements of surface robots in complex environments. Therefore, there is an urgent need for a control method capable of achieving high-precision trajectory tracking in complex flow fields.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] The purpose of this application is to provide a high-precision trajectory tracking control method and related equipment for water surface robots, aiming to solve the problems of traditional environmental perception algorithms being susceptible to interference from mirror reflections, path planning not fully considering the fluid dynamic coupling effect, and trajectory tracking control responding to complex hydrodynamic disturbances in complex water surface environments.

[0008] In a first aspect, this application provides a high-precision trajectory tracking control method for a surface robot, the method comprising the following steps:

[0009] A1. Environmental information is collected through multi-source heterogeneous sensors to fuse and construct a dynamic multimodal environmental map that includes three-dimensional geometric topology information of the water surface and vector information of the water flow field;

[0010] A2. Based on the dynamic multimodal environment map, an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints is used for global path planning, and the planned global path is smoothed and parameterized to generate a reference trajectory.

[0011] A3. Based on the water flow field vector information in the dynamic multimodal environment map, predict the flow field disturbance based on the fluid-structure interaction dynamics model, estimate the system residual disturbance using the extended state observer, and generate the feedforward control quantity by combining the flow field disturbance and the system residual disturbance;

[0012] A4. Based on the real-time water flow velocity and trajectory tracking error, the nonlinear model predictive controller and the adaptive sliding mode controller are adaptively switched. Combined with the feedforward control quantity, control commands are generated to drive the water surface robot to track the reference trajectory.

[0013] Preferably, the multi-source heterogeneous sensor includes a binocular vision camera, a millimeter-wave radar, and an acoustic Doppler current profiler array;

[0014] Secondly, this application provides a high-precision trajectory tracking control system for a surface robot, the system comprising:

[0015] The multimodal environment map generation module is used to collect environmental information through multi-source heterogeneous sensors, and to fuse and construct a dynamic multimodal environment map that includes three-dimensional geometric topology information of the water surface and vector information of the water flow field.

[0016] The reference trajectory generation module is used to perform global path planning based on the dynamic multimodal environment map, using an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints, and to perform smooth parameterization processing on the planned global path to generate a reference trajectory.

[0017] The feedforward control quantity generation module is used to predict the flow field disturbance based on the flow field vector information in the dynamic multimodal environment map and the fluid-structure interaction dynamics model, and to estimate the system residual disturbance using the extended state observer, and to generate the feedforward control quantity by combining the flow field disturbance and the system residual disturbance.

[0018] The control drive module is used to adaptively switch between the nonlinear model predictive controller and the adaptive sliding mode controller based on the real-time water flow velocity and trajectory tracking error, and generate control commands in combination with the feedforward control quantity to drive the water surface robot to track the reference trajectory.

[0019] Thirdly, this application provides an electronic device, including a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it performs the steps of the high-precision trajectory tracking control method for a surface robot as described above.

[0020] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the high-precision trajectory tracking control method for a surface robot as described above.

[0021] Beneficial Effects: This application provides a high-precision trajectory tracking control method and related equipment for a surface robot. It constructs a dynamic multimodal environment map containing three-dimensional geometric topology information of the water surface and water flow field vector information through the fusion of multi-source heterogeneous sensor data, effectively solving the problem of inaccurate environmental perception caused by water surface reflection interference in existing technologies. Based on this, this application employs an improved quantum particle swarm optimization algorithm with embedded water flow dynamics equation constraints for global path planning, and performs smooth parameterization processing on the planned global path to generate a reference trajectory. This overcomes the problem of high energy consumption and difficulty in accurate tracking caused by traditional path planning methods that do not fully consider the coupling effect between fluid and robot dynamics. Furthermore, based on the water flow field vector information in the dynamic multimodal environment map, this application predicts flow field disturbances based on a fluid-structure interaction dynamics model and estimates system residual disturbances using an extended state observer. The flow field disturbances and system residual disturbances are combined to generate feedforward control quantities, effectively compensating for external disturbances. Furthermore, this application adaptively switches between a nonlinear model predictive controller and an adaptive sliding mode controller based on real-time water flow velocity and trajectory tracking error. Combined with feedforward control inputs, it generates control commands to drive the surface robot to track a reference trajectory. This solves the problems of lag and large trajectory tracking errors in traditional controllers when facing complex hydrodynamic disturbances such as nonlinear surges and abrupt flow fields. In summary, this application, through the above technical solution, significantly improves the surface robot's environmental perception capability, path planning efficiency and safety in complex water environments, as well as the accuracy and robustness of trajectory tracking, achieving high-precision trajectory tracking control. It demonstrates significant technological advancement and practical value. Attached Figure Description

[0022] Figure 1 A flowchart of a high-precision trajectory tracking control method for a surface robot provided in this application.

[0023] Figure 2 This is a schematic diagram of a high-precision trajectory tracking control system for a surface robot provided in this application.

[0024] Figure 3 A schematic diagram of the structure of the electronic device provided in this application.

[0025] Labeling Explanation: 1. Multimodal Environment Map Generation Module; 2. Reference Trajectory Generation Module; 3. Feedforward Control Variable Generation Module; 4. Control Drive Module; 301. Processor; 302. Memory; 303. Communication Bus. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0027] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0028] Please refer to Figure 1 This application discloses a high-precision trajectory tracking control method for a surface robot in some embodiments, the method comprising the following steps:

[0029] A1. Environmental information is collected through multi-source heterogeneous sensors to fuse and construct a dynamic multimodal environmental map that includes three-dimensional geometric topology information of the water surface and vector information of the water flow field;

[0030] A2. Based on the dynamic multimodal environment map, an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints is used for global path planning, and the planned global path is smoothed and parameterized to generate a reference trajectory.

[0031] A3. Based on the water flow field vector information in the dynamic multimodal environment map, predict the flow field disturbance based on the fluid-structure interaction dynamics model, estimate the system residual disturbance using the extended state observer, and generate the feedforward control quantity by combining the flow field disturbance and the system residual disturbance.

[0032] A4. Based on the real-time water flow velocity and trajectory tracking error, the nonlinear model predictive controller and the adaptive sliding mode controller are adaptively switched. Combined with the feedforward control quantity, control commands are generated to drive the water surface robot to track the reference trajectory.

[0033] This application aims to provide a high-precision trajectory tracking control method for surface robots, addressing the limitations of existing technologies in environmental perception, path planning, and trajectory tracking control of surface robots in complex aquatic environments. This method utilizes multi-source heterogeneous sensor fusion technology to construct a dynamic multimodal environmental map, providing comprehensive and accurate environmental information for subsequent path planning and trajectory tracking. In the path planning stage, an improved quantum particle swarm optimization algorithm constrained by hydrodynamic equations is introduced to ensure the energy efficiency and safety of the planned path. In the trajectory tracking control stage, feedforward control variables are generated by combining flow field disturbance prediction and system residual disturbance estimation, and an adaptive switching controller effectively addresses complex hydrodynamic disturbances, thereby achieving high-precision trajectory tracking for the surface robot.

[0034] The high-precision trajectory tracking control method for surface robots proposed in this application is based on a series of collaborative steps to ensure the autonomous operation capability and trajectory tracking accuracy of surface robots in complex aquatic environments.

[0035] First, in step A1, environmental information is collected through multi-source heterogeneous sensors to fuse and construct a dynamic multimodal environmental map containing three-dimensional geometric topology information of the water surface and vector information of the water flow field. The "multi-source heterogeneous sensors" here can include, but are not limited to, binocular vision cameras, millimeter-wave radar, and acoustic Doppler current profiler arrays. For example, a binocular vision camera can be used to acquire water surface images, millimeter-wave radar can acquire radar point cloud data of the water surface, and an acoustic Doppler current profiler array can acquire vector information of the water flow field. The raw data collected by these sensors needs to be preprocessed and fused to generate a comprehensive environmental map. For example, the water surface image may need to be polarized filtered to eliminate specular reflection interference, and then image processing algorithms (such as the improved YOLOv7 algorithm) are used to detect floating objects on the water surface to obtain obstacle detection information. Simultaneously, based on the binocular vision images, a binocular vision SLAM algorithm can be used to generate binocular SLAM point cloud data. Finally, obstacle detection information, binocular SLAM point cloud data, radar point cloud data, and water flow vector information are fused to construct a dynamic multimodal environment map that includes three-dimensional geometric topological information of the water surface (such as the location, shape, and size of obstacles) and water flow vector information (such as the flow velocity and direction of water). This environment map is dynamic, meaning it can be updated in real time to reflect changes in the water surface environment.

[0036] Secondly, in step A2, based on the dynamic multimodal environment map, an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints is used for global path planning. The planned global path is then smoothed and parameterized to generate a reference trajectory. Traditional geometric path planning methods often neglect the influence of water flow on the surface robot during path planning. To address this issue, this application introduces an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints. For example, the discretized Navier-Stokes equations can be embedded as physical constraints in the fitness function of the quantum particle swarm optimization algorithm to optimize the energy consumption and safety of the path. The water flow field vector information can include the velocity and direction data of the gridded water flow field. By acquiring the vertical profile distribution data of the water flow, a discretized model of the Navier-Stokes equations for the local flow domain is established, thereby providing the dynamic parameter information of the flow field. This method enables the planned path to better match the motion characteristics of the surface robot in the actual water flow environment, reducing energy consumption and improving safety. The planned global path is usually discrete; to enable the surface robot to track smoothly, it needs to be smoothed and parameterized. For example, B-spline curves can be used to smooth and parameterize the global path obtained from the planning, generating a continuous and smooth reference trajectory.

[0037] Next, in step A3, based on the water flow field vector information in the dynamic multimodal environment map, flow field disturbances are predicted using a fluid-structure interaction (FSI) dynamics model, and system residual disturbances are estimated using an extended state observer. The flow field disturbances and system residual disturbances are then combined to generate a feedforward control quantity. The water flow field significantly disturbs the motion of the surface robot. To improve trajectory tracking accuracy, these disturbances need to be predicted and compensated for. The FSI dynamics model can be used to predict the torques and forces generated by the water flow field on the surface robot, thus obtaining the flow field disturbances. Simultaneously, due to unmodeled dynamics, parameter uncertainties, and unknown external disturbances within the system, these system residual disturbances can be estimated using an extended state observer (using an extended state observer for system residual disturbance estimation is existing technology and will not be detailed here). The predicted flow field disturbances and estimated system residual disturbances are superimposed to generate the feedforward control quantity. This feedforward control quantity can compensate in advance for the influence of water flow and system internal uncertainties on the surface robot's motion, thereby improving the control system's response speed and tracking accuracy. The fluid-structure interaction dynamics model can be a discretized Navier-Stokes equation, which can specifically be described by the following equations based on the immersed boundary method (IBM) to describe the robot-fluid coupling relationship:

[0038] ;

[0039] in, The density of the fluid (water). The dynamic viscosity of the fluid. For flow rate, For fluid pressure, The interaction force between the flow field and the surface robot. This represents the system state (as a vector). Represents the system state at time t. Under certain conditions, the interaction force between the flow field and the surface robot is denoted as N, where N is the number of boundary points between the surface robot and the fluid. For the force acting at the k-th boundary point, For the gridded scale parameters of the water flow field, Let be the position vector of the k-th boundary point at time t.

[0040] Finally, in step A4, based on the real-time water flow velocity and trajectory tracking error, the nonlinear model predictive controller and the adaptive sliding mode controller are adaptively switched. Combined with feedforward control inputs, control commands are generated to drive the surface robot to track the reference trajectory. To cope with the complex and ever-changing surface environment, this application employs an adaptive switching control strategy. For example, the water flow velocity and trajectory tracking error can be acquired in real time. When the water flow velocity is not greater than a preset velocity threshold and the trajectory tracking error is not greater than a preset error threshold (e.g., the preset error threshold is 0.3 m, and the preset velocity threshold is 2 m / s), the nonlinear model predictive controller is selected as the feedback controller. The nonlinear model predictive controller can generate feedback control inputs by optimizing the optimal control sequence within the control time window based on the current system state and the expected system state sequence of the reference trajectory within the prediction time window. Preferably, in solving the optimal control sequence, based on differential flatness theory, the nonlinear system corresponding to the dynamic model is mapped to a flat output space, and a linear prediction model is constructed within the flat output space. This maps the system state with respect to the 6-dimensional state space (the 6 dimensions being longitudinal and transverse positions, lateral position, heading angle, velocity, angular velocity, and acceleration) to a 2-dimensional flat output (the 2 dimensions being longitudinal position and heading angle), thereby significantly reducing computational overhead and greatly improving computational efficiency. When the water flow velocity exceeds a preset velocity threshold or the trajectory tracking error exceeds a preset error threshold, an adaptive sliding mode controller is selected as the feedback controller. The adaptive sliding mode controller can generate feedback control quantities through the sliding mode control model based on the tracking error and its first derivative. Its characteristic is strong robustness to system parameter changes and external disturbances. The feedback control quantity generated by the selected feedback controller is superimposed with the feedforward control quantity generated in step A3 to generate the final control command. These control commands are then sent to the actuators of the surface robot (e.g., vector thrusters) to drive the surface robot to accurately track the reference trajectory.

[0041] The high-precision trajectory tracking control method for surface robots proposed in this application constructs a dynamic multimodal environment map containing three-dimensional geometric topology information of the water surface and vector information of the water flow field through multi-source heterogeneous sensor fusion technology. This innovation overcomes the limitations of traditional single sensors in sensing complex water environments, particularly addressing the interference of water surface reflection on visual detection and the difficulty in effectively identifying dynamic floating objects. By fusing data from binocular vision, millimeter-wave radar, and an acoustic Doppler current profiler array, this application provides more comprehensive and accurate environmental information, laying a solid foundation for subsequent path planning and trajectory tracking.

[0042] In terms of path planning, this application employs an improved quantum particle swarm optimization algorithm with embedded constraints from the hydrodynamic equations for global path planning. The planned global path is then smoothed and parameterized to generate a reference trajectory. Compared to traditional path planning methods that do not fully consider the coupling effect between fluid dynamics and the robot's own dynamics, this application embeds discretized Navier-Stokes equations as physical constraints into the optimization algorithm. This ensures that the planned path considers not only geometric feasibility but also energy consumption and safety. This significantly improves the executability and efficiency of the path in real-world hydrodynamic environments, solving the problems of high energy consumption and difficulty in accurate tracking in complex hydrodynamic fields encountered by traditional methods.

[0043] In terms of trajectory tracking control, this application predicts flow field disturbances based on the water flow field vector information in a dynamic multimodal environment map and a fluid-structure interaction dynamics model. It also estimates system residual disturbances using an extended state observer and generates feedforward control inputs by combining the flow field disturbances and system residual disturbances. Simultaneously, based on real-time water flow velocity and trajectory tracking error, it adaptively switches between a nonlinear model predictive controller and an adaptive sliding mode controller, combining the feedforward control inputs to generate control commands to drive the surface robot to track a reference trajectory. The innovation of this control strategy lies in its combination of feedforward control and adaptive switching feedback control. Feedforward control can compensate for predictable flow field disturbances and system residual disturbances in advance, significantly improving the response speed of the control system. The adaptive switching controller can dynamically select the most suitable controller based on real-time environmental conditions (water flow velocity and trajectory tracking error), thus maintaining high-precision tracking performance under different operating conditions. Compared with traditional PID controllers or single MPC controllers, the control method of this application exhibits stronger robustness and higher tracking accuracy when facing complex hydrodynamic disturbances such as nonlinear surges and abrupt flow fields, effectively solving the problems of slow response and large tracking error of traditional controllers.

[0044] In summary, this application has achieved significant technological advancements in three key areas—environmental perception, path planning, and trajectory tracking control—through multi-source heterogeneous sensor fusion, path planning embedded with hydrodynamic constraints, and adaptive switching feedforward-feedback control strategies. This provides an effective solution for high-precision autonomous operation of surface robots in complex aquatic environments.

[0045] In some implementations, the multi-source heterogeneous sensors include binocular vision cameras, millimeter-wave radar, and acoustic Doppler current profiler arrays.

[0046] Step A1 includes:

[0047] A101. Use a binocular vision camera to acquire water surface images, use millimeter-wave radar to acquire radar point cloud data of the water surface, and use an acoustic Doppler current profiler array to acquire water flow field vector information.

[0048] A102. Polarization filtering is applied to the water surface image to obtain a binocular vision image, and the improved YOLOv7 algorithm is used to detect floating objects on the water surface of the binocular vision image to obtain obstacle detection information.

[0049] A103. Based on binocular vision images, a binocular vision SLAM algorithm is used to generate binocular SLAM point cloud data;

[0050] A104. Integrate obstacle detection information, binocular SLAM point cloud data, and radar point cloud data to generate three-dimensional geometric topology information;

[0051] A105. By fusing three-dimensional geometric topology information and water flow field vector information, a dynamic multimodal environment map is obtained.

[0052] Among them, multi-source heterogeneous sensors specifically refer to a combination of various types of sensors used to collect information about the water surface environment, with the aim of obtaining comprehensive and complementary environmental data. Binocular vision cameras can acquire water surface images, providing rich visual information for identifying objects on the water surface and constructing visual models of the environment. Millimeter-wave radar can acquire radar point cloud data of the water surface; its characteristics include strong penetration and minimal impact from environmental factors such as lighting and weather, providing reliable distance and obstacle information. Acoustic Doppler current profiler arrays are specifically used to acquire water flow field vector information, accurately measuring the velocity and direction of water flow, providing fundamental data for hydrodynamic analysis.

[0053] Further, in step A102, polarization filtering is applied to the water surface image to eliminate specular reflection interference. Specular reflection often occurs on the water surface under sunlight, which severely affects the quality of the visual image and subsequent image processing. Polarization filtering effectively suppresses these reflections, making the detection of floating objects on the water surface clearer and more accurate. Based on this, an improved YOLOv7 algorithm is used to detect floating objects on the water surface from the binocular visual image, thereby obtaining obstacle detection information. The improved YOLOv7 algorithm is a highly efficient target detection algorithm; through optimization, it can more accurately and in real-time identify various floating objects on the water surface and mark them as obstacles.

[0054] In step A103, based on the binocular vision image, a binocular visual SLAM algorithm is used to generate binocular SLAM point cloud data. The binocular visual SLAM (Simultaneous Localization and Mapping) algorithm utilizes the parallax information provided by the binocular camera to perform real-time self-localization and construct a 3D map of the environment. The generated binocular SLAM point cloud data contains detailed 3D structural information of the water surface environment.

[0055] In step A104, obstacle detection information, binocular SLAM point cloud data, and radar point cloud data are fused to generate 3D geometric topology information. This fusion process aims to combine the advantages of different sensors; for example, visual information provides texture and semantics, radar information provides accurate distance and robustness, and obstacle detection information provides clear obstacle boundaries, thereby constructing more complete, accurate, and robust 3D geometric topology information of the water surface.

[0056] Finally, in step A105, the three-dimensional geometric topology information and the water flow field vector information are fused to obtain the dynamic multimodal environment map. This step combines static geometric structure information with dynamic water flow field information to form a comprehensive environmental representation, providing a comprehensive environmental perception foundation for the path planning and control of the surface robot.

[0057] This application's solution employs a multi-source heterogeneous sensor architecture, including a binocular vision camera, millimeter-wave radar, and an acoustic Doppler current profiler array, to acquire water surface environmental information from different dimensions. Specifically, the binocular vision camera, combined with polarization filtering and an improved YOLOv7 algorithm, effectively solves the problem of water surface specular reflection interference, improving the accuracy of floating object detection. The millimeter-wave radar provides reliable obstacle point cloud data under complex lighting or adverse weather conditions. The acoustic Doppler current profiler array directly acquires water flow field vector information, providing accurate data for hydrodynamic analysis. These heterogeneous data are fused to construct an environmental model containing three-dimensional geometric topological information of the water surface, and also incorporate water flow field vector information, thus forming a dynamic multimodal environmental map. This map comprehensively reflects the static obstacle distribution and dynamic water flow characteristics of the water surface environment, providing a solid data foundation for subsequent global path planning and flow field disturbance prediction.

[0058] Through the above technical solutions, this application can significantly improve the comprehensiveness and accuracy of environmental information collection. The collaborative work of multi-source heterogeneous sensors effectively compensates for the limitations of a single sensor in specific environments. For example, polarization filtering technology effectively suppresses the interference of water surface specular reflection on visual detection, making the detection of floating objects on the water surface more accurate. At the same time, by integrating three-dimensional geometric topology information and water flow field vector information, a dynamic multimodal environmental map is constructed, enabling the water surface robot not only to perceive static obstacles but also to grasp the dynamic changes of the water flow field in real time. This provides richer and more reliable environmental perception data for high-precision trajectory tracking, thereby improving the adaptability and operational accuracy of the water surface robot in complex aquatic environments.

[0059] Specifically, the improved quantum particle swarm optimization algorithm with embedded constraints from the hydrodynamic equations is as follows:

[0060] In the fitness function of the quantum particle swarm optimization algorithm, the discretized Navier-Stokes equations are embedded as physical constraints to optimize the energy consumption and safety of the path.

[0061] Specifically, quantum particle swarm optimization (QPSO) is a swarm intelligence-based optimization algorithm that finds optimal solutions by simulating the motion of particles in the solution space. The fitness function is a criterion for evaluating the merits of each particle (i.e., potential path solutions); a higher fitness value indicates that the path solution better meets the optimization objective. In this application, the fitness function is designed to consider not only traditional factors such as path length and obstacle avoidance, but more importantly, to embed discretized Navier-Stokes equations as physical constraints. The discretized Navier-Stokes equations are the discrete form of a set of partial differential equations describing fluid motion, accurately reflecting the dynamic characteristics of the water flow field. By embedding them as constraints in the fitness function, the algorithm will forcibly consider the influence of water flow on robot motion during path planning, making the planned path physically more feasible and more consistent with the actual water flow environment. In practical applications, the water flow field vector information includes the velocity and direction data of the gridded water flow field. These data can be used to obtain the vertical profile of the flow field's water velocity through vector synthesis, and a discretized model of the Navier-Stokes equations for the local flow domain can be established based on the obtained vertical profile distribution data of the water flow (e.g., the equations describing the robot-fluid coupling relationship based on the immersed boundary method (IBM) mentioned earlier). This discretized model provides dynamic parameter information of the flow field, enabling accurate prediction of the impact of the water flow on the robot's motion, such as thrust and drag, during path planning, thereby optimizing the energy consumption and safety of the path. For example, avoiding upstream navigation or utilizing downstream navigation can significantly reduce energy consumption; avoiding turbulent or dangerous eddy regions can improve navigation safety.

[0062] For example, the fitness function can be:

[0063] ;

[0064] in, For the fitness function value, At the starting time, The end time, For time, The dominant normalized weighting coefficients for the flow field. For energy-constrained priority, the weighting coefficients are normalized first. For safety reasons, the weighting coefficients are normalized. For the speed of the surface robot, For the water flow field velocity, For the kinetic energy of the surface robot, The distance between the surface robot and the obstacle. This is the proportionality coefficient.

[0065] This application's approach embeds discretized Navier-Stokes equations as physical constraints into the fitness function of a quantum particle swarm optimization (QS) algorithm, enabling the path planning process to deeply integrate the dynamic characteristics of the water flow field. Specifically, the Navier-Stokes equations describe the changes in physical quantities such as water velocity and pressure over time and space, thus accurately simulating the forces exerted by the water flow on a surface robot. When these equations are discretized and introduced into the fitness function as constraints, the QS algorithm, when searching for the optimal path, no longer only considers geometric distance or obstacle avoidance, but also evaluates the energy consumption and potential risks of each candidate path under the actual water flow. For example, the algorithm tends to select paths that can utilize water flow propulsion, reduce countercurrent resistance, or avoid high shear forces and strong eddy current regions. It is precisely because of the introduction of this physical constraint that the planned path is not only geometrically feasible, but also possesses optimal energy efficiency and the highest safety in terms of dynamics.

[0066] Through the above technical solution, this application can significantly improve the intelligence and adaptability of global path planning for surface robots. Specifically, because the complex dynamic effects of the water flow field are fully considered during the path planning stage, the generated reference trajectory can better adapt to the actual water flow environment, thereby effectively reducing the energy consumption of the surface robot when performing tasks and extending its endurance. At the same time, by avoiding potentially dangerous water flow areas, such as strong eddies or rapids, the navigation safety of the surface robot in complex water environments is significantly improved. This optimization method based on physical constraints makes the trajectory tracking control of the surface robot not only highly accurate but also more robust and reliable in practical applications.

[0067] Preferably, in step A2, B-spline curves can be used to smooth and parameterize the global path obtained from the planning to generate a reference trajectory.

[0068] Specifically, B-spline curves are a mathematical tool widely used in computer-aided design and computer graphics. Their key feature is their ability to generate smooth curves with continuity of arbitrary order. B-spline curves are defined by a set of control points and node vectors. Their local controllability ensures that modifications to the curve affect only a local area, not the entire curve. Furthermore, B-spline curves possess excellent properties such as convex hull and reduced variation, effectively preventing self-intersections of paths and ensuring that the generated trajectories are smooth and conform to physical constraints. In practical applications, the order of the B-spline curve can be chosen based on the desired smoothness and computational efficiency. For example, cubic B-spline curves are often used in situations requiring C² continuity, providing excellent smoothing results.

[0069] The proposed solution employs B-spline curves for smoothing parameterization, effectively addressing the trajectory discontinuity issues that may arise with traditional smoothing methods. The mathematical properties of B-spline curves guarantee high-order geometric continuity of the generated trajectory; for example, cubic B-spline curves ensure the continuity of the second derivative of the trajectory (i.e., the rate of change of curvature). This high-order continuity is crucial for trajectory tracking in surface robots, as it means the robot can move along the trajectory more smoothly and predictably, avoiding abrupt acceleration, deceleration, or turning caused by sudden trajectory changes. This reduces the impact on the actuators and improves the stability of the control system. Furthermore, the local controllability of B-spline curves allows for local adjustments to the global path by modifying only a few control points without recalculating the entire path. This significantly improves the efficiency of path planning and trajectory generation in dynamic environments or scenarios requiring real-time path updates.

[0070] By employing the aforementioned technical solution and using B-spline curves to smooth and parameterize the global path, a reference trajectory with high-order continuity and good smoothness can be generated. This enables the surface robot to achieve more stable and precise motion control when tracking this reference trajectory, effectively reducing trajectory tracking errors and lowering system oscillations and energy consumption caused by trajectory discontinuities. Consequently, this not only improves the high-precision trajectory tracking performance of the surface robot but also extends the service life of the actuators, enhancing the overall reliability and operational efficiency of the system.

[0071] In some implementations, step A4 includes:

[0072] A401. Real-time acquisition of water flow field velocity and trajectory tracking error;

[0073] A402. When the flow velocity in the water field is not greater than the preset flow velocity threshold and the trajectory tracking error is not greater than the preset error threshold, the nonlinear model predictive controller is selected as the feedback controller; when the flow velocity in the water field is greater than the preset flow velocity threshold or the trajectory tracking error is greater than the preset error threshold, the adaptive sliding mode controller is selected as the feedback controller.

[0074] A403. Generate feedback control quantity using the selected feedback controller;

[0075] A404. Superimpose feedback control and feedforward control quantities to generate control commands;

[0076] A405. The actuator that drives the surface robot according to control commands.

[0077] Specifically, in step A401, real-time acquisition of water flow velocity refers to continuously monitoring and acquiring the current water flow velocity information through sensors (such as an acoustic Doppler current profiler array) mounted on the surface robot. Meanwhile, trajectory tracking error refers to the deviation between the surface robot's current actual position and the corresponding expected position on the reference trajectory. This error can be calculated based on the surface robot's positioning information (such as GPS, inertial measurement unit, etc.) and the reference trajectory.

[0078] Step A402 is the core of this application, defining an adaptive switching strategy for two feedback controllers. Specifically, when the water flow velocity is relatively low and the trajectory tracking error is within an acceptable range, the system is considered to be in a relatively stable operating state, and a nonlinear model predictive controller is selected as the feedback controller. The nonlinear model predictive controller, due to its ability to handle system nonlinearity, constraints, and optimize future control performance, can provide high-precision trajectory tracking under such conditions. Conversely, when the water flow velocity is high or the trajectory tracking error exceeds a preset range, it indicates that the system may face significant external disturbances or internal uncertainties, and an adaptive sliding mode controller is selected as the feedback controller. The adaptive sliding mode controller is known for its strong robustness to system parameter uncertainties and external disturbances, effectively suppressing errors and ensuring stable tracking of the surface robot in harsh environments. For example, the preset error threshold can be set to 0.3m, and the preset flow velocity threshold can be set to 2m / s. These thresholds can be adjusted according to the specific performance requirements of the surface robot and the actual application scenario.

[0079] In practical applications, step A403 refers to invoking the corresponding control algorithm to calculate the current feedback control quantity based on the controller type determined in step A402. This feedback control quantity aims to correct the real-time tracking error of the surface robot.

[0080] Furthermore, in step A404, the feedback control quantity is superimposed with the feedforward control quantity generated in step A3 to form the final control command. The feedforward control quantity is mainly used to compensate for predictable flow field disturbances, while the feedback control quantity is used to correct unmodeled dynamic and residual errors. The combination of the two can provide a more comprehensive and accurate control effect.

[0081] Finally, in step A405, the actuators of the surface robot, such as vector thrusters, are driven according to the generated control commands to achieve precise adjustment of the surface robot's motion state, enabling it to track the reference trajectory with high accuracy.

[0082] This application's solution effectively addresses the challenge of a single controller simultaneously achieving high precision and robustness in complex aquatic environments by introducing an adaptive controller switching mechanism based on water flow velocity and trajectory tracking error. Specifically, under conditions of stable water flow and small tracking errors, the nonlinear model predictive controller can achieve optimal control performance and refined trajectory tracking. However, under challenging conditions of turbulent water flow and large tracking errors, the adaptive sliding mode controller can quickly intervene, leveraging its strong disturbance suppression capabilities to ensure the surface robot does not deviate too far from the predetermined trajectory, maintaining system stability. Furthermore, by superimposing feedback control and feedforward control, the system can not only proactively predict and counteract the effects of known disturbances but also respond to and correct unknown disturbances and model errors in real time, thereby significantly improving the adaptability and accuracy of the overall control system.

[0083] Through the above technical solution, surface robots can achieve high-precision trajectory tracking in a wider range of aquatic environments, effectively overcoming the limitations of traditional control methods in terms of performance degradation under varying water flow conditions. This solution significantly improves the reliability and efficiency of surface robot operations by intelligently selecting the controller best suited to the current environment and performance requirements, making it particularly suitable for scenarios requiring long-term, high-precision operations, such as water quality monitoring and underwater mapping.

[0084] Preferably, in step A403, if the selected feedback controller is a nonlinear model predictive controller, the nonlinear model predictive controller generates the feedback control quantity in the following manner:

[0085] Based on the current system status Given the desired system state sequence within the prediction time window based on the reference trajectory, the optimal control sequence within the control time window is solved using the following formula:

[0086] ;

[0087] in, Let k be the system state at time k. Let k+1 be the system state. To predict the length of the time-domain window, The expected system state at time k in the expected system state sequence within the prediction time window is the reference trajectory. To control the optimal control sequence within the time-domain window, the length of the time-domain window is controlled. Less than (For example, It is 15. (5, but not limited to this). Let k be the control quantity at time k. For the preset parameter matrix, Let Q represent the flow velocity in the water flow field, and let Q and R be preset weight matrices. The minimum control output that the system can provide. This is the maximum control quantity that the system can output. The distance between the surface robot and the obstacle. To establish a safe distance, This is a dynamic model of a surface robot;

[0088] Extract the control quantity corresponding to time k=0 from the optimal control sequence. As the feedback control quantity at the current moment.

[0089] Specifically, system status This refers to the kinematic and dynamic state of the surface robot at time k, which may include information such as its longitudinal position, lateral position, heading angle, velocity, angular velocity, and acceleration. Its purpose is to comprehensively describe the robot's current motion. The length of the prediction time-domain window. The time range within which the model predictive controller predicts system behavior is defined, along with the length of the control time-domain window. This indicates the length of the control quantity sequence that needs to be optimized in each control cycle, typically... Less than To balance computational complexity and control performance. The desired system state at time k within the desired system state sequence of the reference trajectory within the prediction time-domain window is used to provide the controller with a clear tracking target. Control variable This is the driving force or torque applied to the surface robot at time k, the purpose of which is to change the robot's motion state. Parameter matrix Used to adjust the flow velocity of the water flow field. The degree of influence on the system dynamics model is determined to more accurately predict the robot's behavior under water flow disturbance. The weight matrices Q and R are preset parameters used to balance trajectory tracking error and control energy consumption during the optimization process. Q is usually related to the state error term, and R is related to the control quantity term. The purpose is to avoid excessive control output while ensuring tracking accuracy. and These represent the minimum and maximum control quantities that the system actuator can output, respectively, with the aim of ensuring that control commands are within physically feasible limits. The distance between the surface robot and the obstacle. These two parameters together constitute the obstacle avoidance constraint, which aims to ensure that the robot avoids collisions with obstacles in the environment while tracking the trajectory. This is a dynamic model for a surface robot, designed to describe the motion of the robot under the influence of control variables and external disturbances.

[0090] The proposed solution generates feedback control input by constructing an optimization problem. This problem aims to minimize the tracking error between the surface robot and the reference trajectory, as well as the cost of control input, while satisfying system dynamics constraints, actuator limitations, and obstacle avoidance safety constraints. Preferably, in solving the optimal control sequence, the nonlinear dynamic model of the surface robot can be mapped to a simplified flat output space based on differential flatness theory. Specifically, the original complex 6-dimensional state space (including longitudinal position, lateral position, heading angle, velocity, angular velocity, and acceleration) is mapped to a 2-dimensional flat output (including only longitudinal position and heading angle). This mapping allows a linear predictive model to be constructed within the flat output space, transforming the original complex nonlinear optimization problem into a computationally more efficient linear or quadratic programming problem. This significantly reduces the computational overhead required to solve the optimal control sequence, enabling the nonlinear model predictive controller to achieve real-time, efficient feedback control on the surface robot. Finally, the control input corresponding to the current time k=0 is extracted from the solved optimal control sequence. This serves as the current feedback control variable, and the process is repeated continuously in a rolling time domain to adapt to environmental changes and system disturbances.

[0091] Through the above technical solution, this application can significantly reduce the computational complexity of nonlinear model predictive controllers in the trajectory tracking application of surface robots, thereby greatly improving the computational efficiency of the control system. This enables surface robots to perform trajectory tracking with faster control cycles and higher response speeds in complex aquatic environments with high real-time requirements. Furthermore, by mapping the nonlinear system to a flat output space and constructing a linear predictive model, not only is control performance guaranteed, but the demand for computing hardware resources is also reduced, improving the robustness and practicality of the system. As a result, surface robots can achieve more accurate and stable trajectory tracking, effectively cope with water flow disturbances and environmental changes, thereby improving their operational efficiency and safety.

[0092] Preferably, in step A403, if the selected feedback controller is an adaptive sliding mode controller, the adaptive sliding mode controller generates the feedback control quantity in the following manner:

[0093] Based on the tracking error, obtain the first derivative of the tracking error;

[0094] Generate feedback control variables based on the following control model:

[0095] ;

[0096] ;

[0097] ;

[0098] in, For feedback control, For sliding surface, and for tracking error, The first derivative of the tracking error, These are the integral weighting coefficients. The boundary layer thickness (e.g., using a preset value of 0.1). and To preset the sliding mode controller parameters, For adaptive sliding mode controller parameters, This is the starting point of the current rolling control cycle. The current time (the rolling control cycle is a control cycle of preset duration. For a rolling control cycle of preset duration T, the start time of the first rolling control cycle is time 0 and the end time is time T. Whenever the time reaches the end point of the current rolling control cycle, the next rolling control cycle is entered, and the start time of the next rolling control cycle is set to the current time. The preset duration T can be set to 10s-15s).

[0099] Specifically, the tracking error *e* refers to the deviation between the current position of the surface robot and the desired position on the reference trajectory, and its first derivative *e* reflects the rate of change of this error over time. The first derivative of the tracking error can typically be obtained by performing differential operations on the real-time tracking error or by estimating it using a state observer and filter. The sliding surface *s* is a core concept in sliding mode control theory. It defines a hyperplane in the system's state space. When the system state is driven onto this hyperplane and slides along it, the system will exhibit the desired dynamic characteristics. In this application, the sliding surface *s* is defined as a linear combination of the tracking error *e* and its integral term, where λ is the integral weighting coefficient used to adjust the influence of the integral term on the formation of the sliding surface, thereby helping to eliminate steady-state errors. Feedback control quantity. The generative model incorporates a proportional term. Differential term and sliding mode control items .in, and These are preset sliding mode controller parameters used to adjust the response characteristics of proportional and derivative control. These are the parameters of the adaptive sliding mode controller, whose values ​​are adaptively adjusted via a function. This adaptive mechanism allows the controller to dynamically adjust the control gain based on the magnitude of the current tracking error e. When the error is large, Increased size provides stronger control to quickly reduce errors; when the error is small... Reduce the frequency to avoid excessive oscillation. The boundary layer thickness is used to introduce a continuous control law near the sliding surface to reduce chattering in traditional sliding mode control and improve the smoothness of control.

[0100] This application's solution effectively addresses the strong disturbances and uncertainties faced by the system when the water flow velocity is high or the trajectory tracking error is large by introducing an adaptive sliding mode controller. When the surface robot is under these challenging conditions, the adaptive sliding mode controller is activated. This controller first constructs a sliding surface s based on the real-time tracking error e and its first derivative e, which aims to guide the system state to converge rapidly to the desired trajectory. Through adaptive adjustment... The parameters allow the controller to dynamically adjust the control strength based on the error magnitude, providing strong correction when the error is large and maintaining stable control when the error is small. Additionally, the boundary layer thickness... The introduction of this mechanism effectively alleviates the chattering problem inherent in traditional sliding mode control, resulting in smoother control output and reduced wear on the actuators. This mechanism enables the surface robot to maintain robust tracking capability of the reference trajectory even in complex and variable water flow environments.

[0101] Through the above technical solution, this application provides a robust and efficient feedback control strategy for surface robots facing challenging conditions such as high-velocity water flow disturbances or large trajectory tracking errors. The adaptive sliding mode controller, through its strong suppression of uncertainties and external disturbances, significantly improves the trajectory tracking accuracy and stability of surface robots in complex environments. In particular, The adaptive adjustment mechanism enables the controller to dynamically optimize its performance based on actual operating conditions, avoiding the performance degradation problem of fixed-parameter controllers under different operating conditions. Meanwhile, the boundary layer thickness... The application of this technology effectively suppressed control chattering, extended the service life of the actuator, and improved the smooth operation performance of the system.

[0102] refer to Figure 2 This application provides a high-precision trajectory tracking control system for a surface robot, the system comprising:

[0103] The multimodal environment map generation module 1 is used to collect environmental information through multi-source heterogeneous sensors, and to fuse and construct a dynamic multimodal environment map containing three-dimensional geometric topology information of the water surface and vector information of the water flow field (for details, please refer to step A1 above).

[0104] Reference trajectory generation module 2 is used to perform global path planning based on the dynamic multimodal environment map, using an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints, and to perform smooth parameterization processing on the planned global path to generate a reference trajectory (the specific process can be referred to step A2 above).

[0105] The feedforward control quantity generation module 3 is used to predict the flow field disturbance based on the flow field vector information in the dynamic multimodal environment map and the fluid-structure interaction dynamic model, and to estimate the system residual disturbance using the extended state observer. The feedforward control quantity is generated by combining the flow field disturbance and the system residual disturbance (the specific process can be referred to step A3 above).

[0106] The control drive module 4 is used to adaptively switch between the nonlinear model predictive controller and the adaptive sliding mode controller based on the real-time water flow field velocity and trajectory tracking error. Combined with the feedforward control quantity, it generates control commands to drive the water surface robot to track the reference trajectory (the specific process can be referred to step A4 above).

[0107] Please refer to Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device includes a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other via a communication bus 303 and / or other forms of connection mechanisms (not shown). The memory 302 stores a computer program executable by the processor 301. When the electronic device is running, the processor 301 executes the computer program to perform the high-precision trajectory tracking control method for a surface robot in any optional implementation of the above embodiments, to achieve the following functions: collecting environmental information through multi-source heterogeneous sensors to fuse and construct a system including three-dimensional geometric topology information of the water surface and water flow field vector information. A dynamic multimodal environment map is generated. Based on the dynamic multimodal environment map, an improved quantum particle swarm optimization algorithm with embedded hydrodynamic equation constraints is used for global path planning. The planned global path is then smoothed and parameterized to generate a reference trajectory. Based on the water flow field vector information in the dynamic multimodal environment map, flow field disturbances are predicted based on a fluid-structure interaction dynamics model. An extended state observer is used to estimate the system residual disturbances. The flow field disturbances and system residual disturbances are combined to generate feedforward control quantities. Based on the real-time water flow velocity and trajectory tracking error, a nonlinear model predictive controller and an adaptive sliding mode controller are adaptively switched. Combined with the feedforward control quantities, control commands are generated to drive the surface robot to track the reference trajectory.

[0108] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it executes the high-precision trajectory tracking control method for a surface robot in any optional implementation of the above embodiments to achieve the following functions: collecting environmental information through multi-source heterogeneous sensors to fuse and construct a dynamic multimodal environment map containing three-dimensional geometric topology information of the water surface and water flow field vector information; performing global path planning using an improved quantum particle swarm optimization algorithm with embedded water flow dynamics equation constraints based on the dynamic multimodal environment map, and performing smooth parameterization processing on the planned global path to generate a reference trajectory; predicting flow field disturbances based on a fluid-structure interaction dynamics model based on the water flow field vector information in the dynamic multimodal environment map, estimating system residual disturbances using an extended state observer, and generating feedforward control quantities by combining flow field disturbances and system residual disturbances; and adaptively switching between a nonlinear model predictive controller and an adaptive sliding mode controller based on real-time water flow velocity and trajectory tracking error, and generating control commands to drive the surface robot to track the reference trajectory by combining the feedforward control quantities. The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0109] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A high-precision trajectory tracking control method for a water surface robot, characterized in that, The method comprises the following steps: A1. Collecting environmental information by a multi-source heterogeneous sensor to fuse and construct a dynamic multi-modal environment map containing three-dimensional geometric topological information and water flow field vector information of the water surface; A2. According to the dynamic multi-modal environment map, a global path planning is performed by using an improved quantum particle swarm optimization algorithm embedded with water flow dynamics equation constraints, and a reference trajectory is generated by performing smoothing parameterization processing on the global path obtained by planning; A3. According to the water flow field vector information in the dynamic multi-modal environment map, a flow field disturbance is predicted based on a fluid-structure coupling dynamics model, a system residual disturbance is estimated by using an extended state observer, and a feedforward control amount is generated in combination with the flow field disturbance and the system residual disturbance; A4. According to the real-time water flow field flow rate and trajectory tracking error, an adaptive switching is performed between a nonlinear model predictive controller and an adaptive sliding mode controller, a control instruction is generated in combination with the feedforward control amount to drive the water surface robot to track the reference trajectory; The improved quantum particle swarm optimization algorithm embedded with water flow dynamics equation constraints is as follows: In the fitness function of the quantum particle swarm optimization algorithm, the discretized Navier-Stokes equation is embedded as a physical constraint condition to optimize the energy consumption and safety of the path; Step A4 comprises: A401. Real-time acquisition of water flow field flow rate and trajectory tracking error; A402. When the water flow field flow rate is not greater than a preset flow rate threshold and the trajectory tracking error is not greater than a preset error threshold, a nonlinear model predictive controller is selected as a feedback controller; when the water flow field flow rate is greater than the preset flow rate threshold or the trajectory tracking error is greater than the preset error threshold, an adaptive sliding mode controller is selected as a feedback controller; A403. Generating a feedback control amount by using the selected feedback controller; A404. Superimposing the feedback control amount and the feedforward control amount to generate a control instruction; A405. Driving the actuator of the water surface robot according to the control instruction.

2. The high-precision trajectory tracking control method for water surface robots according to claim 1, characterized in that, The multi-source heterogeneous sensor comprises a binocular vision camera, a millimeter wave radar and an acoustic Doppler current profiler array; Step A1 comprises: A101. Collecting water surface images by using the binocular vision camera, acquiring radar point cloud data of the water surface by using the millimeter wave radar, and acquiring water flow field vector information by using the acoustic Doppler current profiler array; A102. Polarization filtering the water surface images to obtain binocular vision images, and performing water surface floating object detection on the binocular vision images by using an improved YOLOv7 algorithm to obtain obstacle detection information; A103. Based on the binocular vision images, a binocular vision SLAM algorithm is used to generate binocular SLAM point cloud data; A104. Fusing the obstacle detection information, the binocular SLAM point cloud data and the radar point cloud data to generate three-dimensional geometric topological information; A105. Fusing the three-dimensional geometric topological information and the water flow field vector information to obtain the dynamic multi-modal environment map.

3. The high-precision trajectory tracking control method for water surface robots according to claim 1, characterized in that, In step A2, a B-spline curve is used to perform smoothing parameterization processing on the global path obtained by planning to generate a reference trajectory.

4. The high-precision trajectory tracking control method for water surface robots according to claim 1, characterized in that, In step A403, if the selected feedback controller is a nonlinear model predictive controller, the nonlinear model predictive controller generates a feedback control amount by the following way: According to the current system state and the sequence of desired system states of the reference trajectory within the prediction time domain window, the optimal control sequence within the control time domain window is solved by the following formula: ； wherein, is the system state at time k, is the system state at time k+1, is the length of the prediction horizon, is the desired system state at time k in the sequence of desired system states of the reference trajectory within the prediction horizon, is the optimal control sequence within the control horizon, the length of the control horizon is the control amount at time k, is a preset parameter matrix, is the flow velocity of the water flow field, Q and R are both preset weight matrices, is the minimum control amount that the system can output, and is the maximum control amount that the system can output, is the distance between the water surface robot and the obstacle, is a preset safety distance, is the dynamics model of the water surface robot; Extract the control amount corresponding to k = 0 in the optimal control sequence as the feedback control amount of the current time.

5. The high-precision trajectory tracking control method for water surface robots according to claim 1, characterized in that, In step A403, if the selected feedback controller is an adaptive sliding mode controller, the adaptive sliding mode controller generates a feedback control amount by the following way: According to the tracking error, obtain the first derivative of the tracking error; According to the following control model, generate a feedback control amount: ； ； ； wherein, is the feedback control amount, is a sliding surface, is the tracking error, is a first derivative of the tracking error, is an integral weight coefficient, is a boundary layer thickness, and is a preset sliding mode controller parameter, is an adaptive sliding mode controller parameter, is a time start point of a current rolling control period, is a current time.

6. A high-precision trajectory tracking control system for a water surface robot, characterized in that, The system comprises: A multi-modal environment map generation module is configured to collect environment information through a plurality of heterogeneous sensors, and to fuse and construct a dynamic multi-modal environment map comprising three-dimensional geometric topological information and water flow field vector information of a water surface; A reference trajectory generation module is configured to perform global path planning by using an improved quantum particle swarm optimization algorithm with embedded water flow dynamics equation constraints, and to perform smoothing parameterization processing on the global path obtained by planning, to generate a reference trajectory, according to the dynamic multi-modal environment map; A feedforward control amount generation module is configured to predict flow field disturbance based on a fluid-structure coupling dynamics model according to water flow field vector information in the dynamic multi-modal environment map, to estimate system residual disturbance by using an extended state observer, and to generate a feedforward control amount in combination with the flow field disturbance and the system residual disturbance; A control driving module is configured to adaptively switch a nonlinear model predictive controller and an adaptive sliding mode controller according to a real-time water flow field flow rate and a trajectory tracking error, to generate a control instruction in combination with the feedforward control amount, to drive a water surface robot to track the reference trajectory. The improved quantum particle swarm optimization algorithm with embedded water flow dynamics equation constraints is as follows: In the fitness function of the quantum particle swarm optimization algorithm, a discretized Navier-Stokes equation is embedded as a physical constraint condition, to optimize energy consumption and safety of the path. When the control driving module adaptively switches the nonlinear model predictive controller and the adaptive sliding mode controller according to the real-time water flow field flow rate and the trajectory tracking error, and generates the control instruction in combination with the feedforward control amount to drive the water surface robot to track the reference trajectory, the following steps are performed: A401. Real-time acquisition of water flow field flow rate and trajectory tracking error; A402. When the water flow field flow rate is not greater than a preset flow rate threshold and the trajectory tracking error is not greater than a preset error threshold, a nonlinear model predictive controller is selected as a feedback controller; when the water flow field flow rate is greater than the preset flow rate threshold or the trajectory tracking error is greater than the preset error threshold, an adaptive sliding mode controller is selected as a feedback controller; A403. Generation of a feedback control amount by using the selected feedback controller; A404. Superposition of the feedback control amount and the feedforward control amount to generate a control instruction; A405. Driving of an execution mechanism of the water surface robot according to the control instruction.

7. An electronic device, comprising: The computer program is executed by the processor to run the steps of the high-precision trajectory tracking control method of the water surface robot according to any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to run the steps of the high-precision trajectory tracking control method of the water surface robot according to any one of claims 1-5.

Images