Unmanned ship distributed cooperative dynamic target surrounding control method under multiple constraints

Abstract

The invention discloses an unmanned ship distributed cooperative dynamic target surrounding control method under multiple constraints, and relates to the field of unmanned ship cooperative control. In order to solve the problems that multiple constraint coupling is not fully considered in cooperative surrounding of multiple unmanned ships, target state estimation is insufficient in an intermittent communication environment and the like, the following technical scheme is provided: firstly, a motion model of the unmanned ships and a target is established; a distributed internal model observer is arranged for each unmanned ship, and the internal motion state of the target is estimated; then designing a distributed type progressive average surrounding guidance algorithm, and performing distributed scheduling on the expected speed and heading of the unmanned ship, so that the unmanned ship can still keep safe surrounding when the speed or direction of the target is changed; the input saturation constraint of a propeller or a steering engine is considered, so that each unmanned ship can be autonomously controlled according to an expected trajectory and speed; and finally, stable and collision-free surrounding of the dynamic target is realized under multiple constraint conditions. The method can be applied to scenes such as ocean escort and maritime search and rescue which need to perform collaborative operation around a dynamic target.

Description

Technical Field

[0001] It relates to the field of cooperative control of unmanned boats. Background Art

[0002] In recent years, unmanned surface vehicles (USVs) have been widely used in scenarios such as marine patrol, marine monitoring, search and rescue, and security escort due to their flexible maneuverability and high mission adaptability. With the increasing demand for marine tasks, the limitations of single-boat mission execution have become more and more obvious; therefore, cooperative operation of multiple unmanned boats has gradually become a key research direction in this field. Cooperative control of multiple unmanned boats has shown significant advantages in distributed monitoring, surface communication networking, cooperative formation, target tracking and surrounding, etc., which can greatly improve mission efficiency and system robustness.

[0003] In terms of cooperative surrounding control, existing research mostly focuses on distributed control:

[0004] Method based on centralized perception and planning:

[0005] Early research usually adopted a centralized architecture, integrating all state information of the unmanned boat group in the command center to uniformly plan the trajectory or control input. Such methods are relatively simple to implement and can theoretically obtain the global optimal solution. However, this architecture relies too much on communication bandwidth and the computing power of the central node. Once the communication is interrupted or the central node fails, the entire system will be difficult to maintain normal cooperation, and the robustness is relatively insufficient.

[0006] For example, in some nearshore defense scenarios, shore-based radars or central servers can provide global situation information for centralized command. However, once the unmanned boat is far from the shore-based base station or executes tasks in a strong interference environment, the communication link may be severely blocked, resulting in the failure of centralized control.

[0007] Method based on distributed consensus and neighbor interaction:

[0008] Subsequent research has tended more towards distributed control, enabling unmanned boats to exchange information with adjacent boats by constructing an adjacency topology to achieve target tracking or surrounding. Among them, significant progress has been made in cooperative formation and target tracking of multiple unmanned boats by distributed consensus algorithms (Distributed Consensus), distributed observers, etc.

[0009] Some work enables unmanned boats to maintain a reasonable formation and have a certain collision avoidance ability when surrounding the target through phase angle repulsion or virtual leader strategies. In such methods, each boat can complete control decisions based on local neighbor information and a small amount of target information, improving the adaptability of the system in scenarios with limited communication.

[0010] Robust / Adaptive Control Against External Interference and Model Uncertainty:

[0011] Facing disturbances such as wind waves and ocean currents in the marine environment and the perturbation of the unmanned surface vehicle (USV) model, some studies have introduced robust control, adaptive control, or intelligent algorithms (such as fuzzy control, neural network approximation, etc.) to improve control accuracy and system stability.

[0012] For example, some literature uses adaptive neural networks to learn the complex hydrodynamic characteristics of USVs, updates control parameters online, and reduces the dependence on accurate mathematical models. There are also studies that combine observers and disturbance rejection techniques to improve the anti-interference performance of the system against random disturbances.

[0013] Control Strategies Considering Input Saturation and Network Constraints:

[0014] USVs generally face the problem of actuator thrust or rudder angle saturation during actual use, especially in high-load or strong current environments. Traditional linear controllers are difficult to ensure effective regulation of the system under saturation conditions, so methods such as input saturation compensation and anti-saturation adaptation have emerged.

[0015] In addition, in the cooperative control of multiple USVs, communication constraints (such as time delay, packet loss, intermittent communication) will significantly increase the difficulty of system design. Some distributed control algorithms attempt to reduce the communication burden by using event-triggered or intermittent exchange strategies, but often fail to consider the coupled effects of multiple factors such as model uncertainty, external interference, and input saturation.

[0016] Although the above studies have made varying degrees of progress in the cooperative control of multiple USVs, there are still the following deficiencies:

[0017] The coupling of multiple constraints is not fully considered: including model parameter uncertainty, external random interference, actuator input saturation, and underactuated characteristics, etc.; when dealing with one or some of the constraints, the influence of other factors is often ignored, resulting in performance degradation or even instability in the actual system.

[0018] Insufficient target state estimation in intermittent communication environments: Distributed control relies on communication topologies and neighbor states, but it is difficult to estimate the target state with high accuracy when communication is discontinuous or there are packet losses, affecting subsequent guidance and control effects.

[0019] Weak real-time adaptability: In a dynamic marine environment, simple offline planning or fixed parameters are difficult to respond quickly to model and environmental perturbations; even when using adaptive methods, there is a lack of robust stability analysis under multiple complex constraints.

[0020] In summary, there are still certain limitations in the existing technologies when dealing with challenges such as multiple coupling constraints, intermittent communication, and environmental interference in the process of multi-unmanned boat collaborative surrounding. Further research and improvement are urgently needed. Based on the above technical bottlenecks and deficiencies, this solution proposes a distributed collaborative dynamic target surrounding control method for unmanned boats under multiple constraints to better balance the safety, stability, and robustness of the system. Summary of the Invention

[0021] To solve the technical problems existing in the prior art, in the existing multi-unmanned boat collaborative surrounding control, the coupling of multiple constraints, the insufficient estimation of the target state in an intermittent communication environment, and the weak real-time adaptation ability are not fully considered. The technical solution provided by the present invention is as follows:

[0022] A distributed collaborative dynamic target surrounding control method for unmanned boats under multiple constraints, including:

[0023] Step 1: Establish a motion model;

[0024] Design a distributed internal model observer:

[0025] In view of the intermittency and instability of the marine communication environment, a distributed internal model observer is set for each unmanned boat, and the internal motion state of the target is estimated through the iterative update of the state data exchange between each other and the observation data of the boat itself.

[0026] Step 2: Design a distributed asymptotically average surrounding guidance algorithm:

[0027] Distribute the scheduling of the desired speed and desired heading of the unmanned boats, so that multiple unmanned boats can still maintain a safe surrounding formation when the speed or direction of the dynamic target changes.

[0028] Step 3: Consider the input saturation constraint and perform adaptive learning dynamic control:

[0029] Compensate for the input saturation of the propellers or rudders of the unmanned boats, and perform real-time approximation and correction on external environmental interference and the uncertainty of the unmanned boat model, so that each unmanned boat is controlled according to the desired trajectory and motion speed.

[0030] Step 4: Implement multi-constraint control:

[0031] Complete the stable and collision-free surrounding of the dynamic target.

[0032] Furthermore, step 1 is specifically: obtain the position information, heading information, and speed information of multiple unmanned boats, as well as the position information and motion state information of the dynamic target, and uniformly represent the kinematic models of the unmanned boats and the target in the same coordinate system.

[0033] Furthermore, according to the target state information output by the distributed internal model observer, a distance-based, angle-based, and phase angle repulsion strategy is constructed to perform distributed scheduling on the desired speed and desired heading of the unmanned surface vehicle.

[0034] Furthermore, a dead zone operator model is used to compensate for the input saturation of the thrusters or rudders of the unmanned surface vehicle. At the same time, combined with an adaptive online learning algorithm, the external environmental disturbances and the uncertainties of the unmanned surface vehicle model are approximated and corrected in real time through neural networks or minimum parameter learning methods, so that each unmanned surface vehicle is controlled according to the desired trajectory and motion speed given by the distributed approximate average enclosing guidance algorithm.

[0035] Furthermore, the output results of the distributed internal model observer, the desired states of the distributed approximate average enclosing guidance algorithm, and the dynamic control process driven by adaptive learning are comprehensively executed to complete the stable and collision-free enclosing of the dynamic target.

[0036] There is also provided a distributed cooperative dynamic target enclosing control device for an unmanned surface vehicle under multiple constraints, including:

[0037] Module 1: Establish a motion model;

[0038] Design a distributed internal model observer:

[0039] In view of the intermittency and instability of the marine communication environment, a distributed internal model observer is set for each unmanned surface vehicle, and the internal motion state of the target is estimated through the iterative update of the state data exchange between each other and the observation data of the vehicle itself.

[0040] Module 2: Design a distributed approximate average enclosing guidance algorithm:

[0041] Perform distributed scheduling on the desired speed and desired heading of the unmanned surface vehicle, so that multiple unmanned surface vehicles can still maintain a safe enclosing formation when the speed or direction of the dynamic target changes.

[0042] Module 3: Dynamics control considering input saturation constraints and performing adaptive learning:

[0043] Compensate for the input saturation of the thrusters or rudders of the unmanned surface vehicle, approximate and correct the external environmental disturbances and the uncertainties of the unmanned surface vehicle model in real time, so that each unmanned surface vehicle is controlled according to the desired trajectory and motion speed.

[0044] Module 4: Implement multi-constraint control:

[0045] Complete the stable and collision-free enclosing of the dynamic target.

[0046] There is also provided a computer storage medium for storing a computing program, and when the computer program is read by a computer, the computer executes the described method.

[0047] A computer is also provided, which includes a processor and a storage medium. When the processor reads the computer program stored in the storage medium, the computer executes the described method.

[0048] A computer program product is also provided. As a computer program, when the computer program is executed, the described method is implemented.

[0049] Compared with the prior art, the beneficial effects of the technical solution provided by the present invention are as follows:

[0050] By introducing a distributed internal model observer in the communication layer, the system significantly improves the accuracy and robustness in estimating the state of unknown or partially unknown targets. Especially in an environment with intermittent communication or packet loss, it can still obtain approximate real target motion information. Compared with the traditional research methods based on global perception or relying on centralized communication, this distributed observer better adapts to the dynamic network characteristics in the marine environment and reduces the failure risk caused by central node failures or communication bottlenecks.

[0051] In the guidance layer, a method similar to the progressive average surrounding method is adopted and combined with the phase angle repulsion theory. This not only ensures that multiple unmanned boats can maintain a sufficient relative distance during the process of surrounding a dynamic target, thus effectively reducing the risk of collision between boats, but also can quickly adjust the surrounding speed and angle when the target's motion speed or direction changes. Compared with the previous methods that only rely on fixed formations or pre-planned curves, this distributed guidance scheme has more advantages in terms of scalability and self-adaptability, and also helps to shorten the response time to changes in the external environment.

[0052] In the control layer, by using the dead zone operator model for actuator input saturation compensation and combining a radial basis neural network for online approximation and adaptive learning, the system can still maintain stable and accurate control performance under various strong interference and nonlinear coupling constraints. Most existing studies only consider linear saturation compensation or only perform robust control for a single uncertainty, while this scheme couples multiple factors such as model uncertainty, external disturbances, and input saturation into a unified adaptive framework, enabling the unmanned boat to obtain reliable control and target tracking effects in a wider range of working conditions.

[0053] Due to the introduction of the minimum parameter learning method and the combination of distributed observation information, the controller can still quickly adjust the control gain when there are large perturbations in unknown parameters or sudden changes in external disturbances, enhancing the real-time performance and anti-interference ability of the system. Compared with the general methods that identify parameters offline or pre-tune fixed gains, the online learning and adaptive adjustment method adopted in this scheme can significantly shorten the parameter convergence time, enabling the unmanned boat swarm to maintain better self-adaptability and accuracy in response to various environmental changes during the task execution process.

[0054] It can be applied to scenarios such as ocean escort and maritime search and rescue that require surrounding and collaborative operations on dynamic targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic flow diagram of a distributed cooperative dynamic target surrounding control method for unmanned boats under multiple constraints;

[0056] Figure 2 It is a curve graph of the change of the target surrounding trajectory;

[0057] Figure 3 It is a curve graph of the change of the distance error;

[0058] Figure 4 It is a curve graph of the change of the angle error. DETAILED IMPLEMENTATION MANNER

[0059] To make the advantages and beneficial effects of the technical solution provided by the present invention more clearly manifested, the technical solution provided by the present invention will be further described in detail below in conjunction with the accompanying drawings. Specifically:

[0060] Embodiment 1. This embodiment provides a distributed cooperative dynamic target surrounding control method for unmanned boats under multiple constraints, including:

[0061] Step 1: Establish a motion model;

[0062] Design a distributed internal model observer:

[0063] In view of the intermittency and instability of the marine communication environment, a distributed internal model observer is set for each unmanned boat, and the internal motion state of the target is estimated through the exchange of state data between each other and the iteration update of the observation data of this boat.

[0064] Step 2: Design a distributed approximate average surrounding guidance algorithm:

[0065] Distribute the scheduling of the desired speed and desired heading of the unmanned boat, so that multiple unmanned boats can still maintain a safe surrounding formation when the speed or direction of the dynamic target changes.

[0066] Step 3: Consider the input saturation constraint and perform adaptive learning dynamic control:

[0067] Compensate for the input saturation of the propeller or rudder of the unmanned boat, and perform real-time approximation and correction on the external environmental interference and the uncertainty of the unmanned boat model, so that each unmanned boat is controlled according to the desired trajectory and motion speed.

[0068] Step 4: Implement multi-constraint control:

[0069] Complete the stable and collision-free surrounding of the dynamic target.

[0070] Step 1 specifically includes: obtaining the position information, heading information, and speed information of multiple target unmanned boats, as well as the position information and motion state information of dynamic targets, and uniformly representing the kinematic models of the unmanned boats and targets in the same coordinate system.

[0071] According to the target state information output by the distributed internal model observer, construct a surrounding distance, surrounding angle, and phase angle repulsion strategy to perform distributed scheduling on the desired speed and desired heading of the unmanned boat.

[0072] Use the dead zone operator model to compensate for the input saturation of the unmanned boat's thruster or rudder, and at the same time, combined with the adaptive online learning algorithm, use neural network or minimum parameter learning methods to approximate and correct the external environmental interference and the uncertainty of the unmanned boat model in real time, so that each unmanned boat is controlled according to the desired trajectory and motion speed given by the distributed class-asymptotic average surrounding guidance algorithm.

[0073] Integrate and execute the output results of the obtained distributed internal model observer, the desired state of the distributed class-asymptotic average surrounding guidance algorithm, and the dynamic control process driven by adaptive learning to complete the stable and collision-free surrounding of the dynamic target.

[0074] Embodiment 2: This embodiment is a further description of the technical solution provided in Embodiment 1. Specifically:

[0075] The first step: Establish the motion models of the unmanned boat and the target

[0076] In this step, it is necessary to establish the motion models of multiple unmanned boats and dynamic targets respectively, and clarify their respective position information, speed information, and heading information. Through unified coordinate system and symbol definitions, accurate data bases are provided for subsequent target state estimation and distributed control design.

[0077] Detailed description:

[0078] First, determine the sea surface coordinate system as the inertial coordinate system to represent the global poses of the unmanned boat and the target. Each unmanned boat contains basic state information such as position coordinates and heading angles, and its motion speed can be decomposed into speed components along the forward and lateral directions, as well as rotational angular velocity, etc.

[0079] Regard the motion of each unmanned boat as a system that can be driven by forces and torques. External environmental disturbances (such as wind waves, ocean currents, etc.) and the uncertainties of the boat body itself may affect the actual dynamic performance. To lay a foundation for subsequent control and observation, it is necessary to parameterize or lump these disturbances and uncertainties conceptually.

[0080] Meanwhile, the target (such as a moving ship or a tracked target) also has changes in its own position and direction, which can also be described as a motion state related to time, and there may be unknown speeds, accelerations, or external interferences. By representing the motion of the target in the same coordinate system, it provides a reference for subsequent observer design and guidance.

[0081] After the above model is established, the preliminary kinematic descriptions of the "unmanned boat and the target" in the same coordinate system can be obtained, and this descriptive information will become the input basis for estimating the target state in the next step.

[0082] Step 2: Design of the distributed internal model observer

[0083] In this step, using the motion information of the unmanned boat and the target established in the first step, a distributed internal model observer is introduced to estimate the internal state of the target. Since the communication environment at sea may be intermittent or packet-lossy, the observer needs to have high robustness and adaptability, and be able to accurately reconstruct the motion information of the target even when partial communication fails.

[0084] Detailed description:

[0085] First, based on the target motion model and the network communication structure between the unmanned boats, a suitable distributed architecture is selected, so that each unmanned boat only needs to exchange data with its neighboring boats to iteratively update the estimation of the target state. This can maintain the overall observability of the system even when the marine network link is unstable.

[0086] The core of the distributed internal model observer is to design a set of internal states for each unmanned boat, combine the known inputs (such as thrust information) with the neighbor observation error signals obtained during mutual communication, and continuously correct and approximate the key quantities such as the true position and speed of the target. If the communication link is temporarily interrupted, the observer can also use the information available at the previous moment of this boat or in the vicinity for short-term prediction to reduce the impact of data loss.

[0087] When the communication resumes, the unmanned boats exchange the observation deviation information with their neighbors again to correct the estimation of the target state by each boat in an overall manner, so that each boat can obtain a relatively consistent and accurate estimation result of the target state.

[0088] Through the distributed internal model observer in this step, the output target state (such as the estimated values of the target position, speed, etc.) will be transmitted to the guidance layer for use.

[0089] Step 3: Distributed asymptotically-average-like surrounding guidance algorithm

[0090] Based on the target state obtained in the previous step, a distributed class-progressive-average surrounding guidance algorithm is designed in this step, enabling multiple unmanned boats to surround a dynamic target without collision and adjust the surrounding trajectory in a timely manner when the target may change speed or direction.

[0091] Detailed description:

[0092] First, based on the target position and velocity information output by the observer, indicators such as surrounding distance and surrounding angle are defined for the unmanned boats. The unmanned boats need to continuously regulate the longitudinal and lateral velocities, steering rates, etc. to achieve the desired "surrounding" form.

[0093] Due to possible mutual interference or safety distance requirements among the unmanned boats, a phase angle repulsion strategy is introduced here, enabling the unmanned boats to maintain a certain relative angle or spacing during the surrounding process to prevent overcrowding or collision of the boat group. The mechanism of class-progressive-average can achieve gradual approximation and surrounding of the dynamic target and make adaptive adjustments as the target speed and direction change.

[0094] The entire guidance algorithm runs in a distributed manner: each boat can continuously update its desired surrounding velocity and desired heading change based on the currently obtained target state information and the phase angle error information of adjacent boats, ensuring that all unmanned boats form a relatively uniform and safe target encirclement.

[0095] After completing this guidance algorithm, the desired motion states of each unmanned boat (such as desired longitudinal velocity, heading angular velocity, etc.) will be sent to the control layer in the next step for execution.

[0096] Step 4: Dynamics control method based on input saturation and adaptive learning

[0097] After obtaining the desired surrounding motion state in the previous step, it is necessary to consider the actual dynamic characteristics of the propulsion system and steering system of the unmanned boat, especially multiple factors such as saturation constraints, external disturbances, and unknown or inaccurate models. In this step, by introducing a dead zone operator model and a radial basis neural network learning mechanism into the control strategy, adaptive compensation for input saturation and rapid approximation of the unknown environment are achieved.

[0098] Detailed description:

[0099] First, controllers are set for each unmanned boat respectively, taking the desired speed command and heading command as input targets. Since the unmanned boats often have "upper limits" or "nonlinear responses" in aspects such as engine or propeller thrust and rudder angle, which will cause deviations between the actual output and the ideal input, it is necessary to use a dead zone operator model inside the controller to compensate for this nonlinear saturation.

[0100] To further enhance the adaptability to ocean environmental disturbances and internal model uncertainties, an online learning part of a radial basis neural network is incorporated into the controller, which adjusts the neural network weights at any time according to the observed velocity error and position deviation, so as to continuously approximate the unmodeled or time-varying disturbance quantities in the system.

[0101] In this process, the minimum parameter learning method can be used to simplify the update process of neural network weights or disturbance estimation parameters, making the design of the controller more practical and stable. Through this control method, ultimately the unmanned boat can maintain sufficient control accuracy and fast response ability in an environment with multiple constraints superimposed (including underactuation, saturation limits, random ocean currents, etc.), so as to follow the desired state output by the guidance layer.

[0102] The output of this control method is the final real-time execution instructions for each unmanned boat, including specific operation signals such as thrust allocation and rudder adjustment, so as to effectively surround the target in the real ocean environment.

[0103] Step 5: Comprehensive implementation process and functional effects

[0104] The above steps form a complete distributed cooperative dynamic target surrounding control process for unmanned boats under multiple constraints: starting from modeling and observation, to the guidance algorithm allocating motion targets, and then to the controller compensating for saturation and disturbances, connecting and cooperating layer by layer, and ultimately achieving efficient cooperative surrounding of dynamic targets in the ocean environment.

[0105] Detailed description:

[0106] When the unmanned boat cluster receives external task instructions or detects a specified target in the operation area, it first shares target information through the aforementioned distributed internal model observer in an intermittent communication network environment, enabling the observer to continuously update the target position and velocity.

[0107] Through the class-progressive average surrounding guidance device and the phase angle repulsion strategy, if the target motion speed or direction changes, each unmanned boat can recalculate and adjust its own desired speed and heading in a timely manner to achieve collision-free and multi-angle cooperative surrounding.

[0108] The controller of each unmanned boat compensates for the saturation effect of the thruster or rudder during the execution process by using the dead zone operator model, and learns the external disturbance characteristics and model perturbation conditions through a radial basis neural network to ensure the overall control accuracy and stability.

[0109] Multiple unmanned boats finally maintain an appropriate formation and relative distance around the target to achieve the expected surrounding effect. This implementation plan can be widely applied to task scenarios such as ocean escort, maritime blockade, target surveillance, and search and rescue, and has good adaptability and robustness.

[0110] Embodiment 3. CombinationFigures 1-4 For this embodiment, this embodiment further defines the above-provided technical solution through specific embodiments:

[0111] (1) Establish a motion mathematical model. Consider a swarm system composed of N unmanned boats. In the geodetic coordinate system, the kinematic model of the i-th unmanned boat is expressed as follows:

[0112]

[0113] In the formula, [x i , y i T represents the position coordinates of the unmanned boat, represents the heading angle of the unmanned boat; u i , v i and r i respectively represent the longitudinal speed, lateral speed and yaw angular velocity of the unmanned boat. Define as the position vector of the i-th unmanned boat, and define B i = [u i , v i , r i T as the speed vector of the i-th unmanned boat.

[0114] The dynamic model of the i-th unmanned boat is:

[0115]

[0116] and

[0117]

[0118] In the formula, and r t i respectively represent the longitudinal speed, lateral speed and yaw angular velocity of the i-th unmanned boat. F1 i and represent the model parameter terms. and represent the lumped disturbance terms, and represent the external disturbances, f1 i and represent the model uncertainty terms. and respectively represent the longitudinal thrust and yaw moment. X * , Y * and N * represent the hydrodynamic coefficients. m represents the mass of the unmanned boat. I zz represents the inertia matrix. ​​

[0119] In the geodetic coordinate system, the kinematic model of the target is expressed as follows:

[0120]

[0121] where [x t , y t T represents the target position coordinates, represents the target heading angle; u t , v t and r t respectively represent the target longitudinal velocity, lateral velocity, and yaw angular velocity. Define as the position vector of the target, and define B t = [u t , v t , r t T as the velocity vector of the target.

[0122] Moreover, the mathematical model of the target can be expressed as follows:

[0123]

[0124] where p0 = [x t , y t T and q0(t) = [u t , v t T respectively represent the position state and velocity state of the target, represents the uncertain local disturbance vector, and m represents an unknown parameter. The real parts of all eigenvalues λ1, λ2,..., λ n of F(m) are all 0, that is, Re(λ i ) = 0, 1 ≤ i ≤ n, where n represents the dimension of the matrix F(m).

[0125] (2) Define the distance between the i-th unmanned boat and the target as Define the distance between the i-th unmanned boat and the j-th unmanned boat as Define the target circumferential angle as where β i represents the angle of the target relative to the i-th unmanned boat. Define the angle of the i-th unmanned boat relative to the target as φ i . The included angle between the i-th unmanned boat and the j-th unmanned boat can be expressed as θ ij = φ i - φ j + 2kπ ∈ (-π, π]. Define the expected angle of the target circumferential as η d_i , and the expected distance of the target circumferential as ρ dThus, the target tracking error of each unmanned boat is η e_i = η i -η d_i and ρ e_i = ρ i -ρ d .

[0126] (3) According to the given target model, design a distributed internal model observer to estimate the target state, which is expressed as follows:

[0127]

[0128] where m i (t) represents the estimated value of m, y i (t) represents the estimated value of the target position p0(t), represents the network communication signal, U i (t) represents the internal model, and k and w represent positive design parameters.

[0129] Considering the characteristics of the variable y i (t), it is divided into a steady-state part and a dynamic part. Define as the steady-state part, and holds. In addition, it can be obtained that holds. The differential form of the steady-state component of y i (t) is:

[0130]

[0131] Then, assume that there exists a sufficiently smooth function s(q0(t), q d ): Its function value at the origin is 0, i.e., s(0, 0) = 0. For all

[0132]

[0133] The Lie derivative of K1(q0, m, q d ) is defined as follows:

[0134]

[0135] Define a steady-state generator as:

[0136]

[0137] K i (q0, m, q d ) = γ i q i (q0, m, q d )

[0138]

[0139] wherein, represents a scalar number, l i represents a positive integer, is observable. Select a pair of controllable matrices (Ψ i , Θ i ), where Ψ i is a Hurwitz matrix. Since the matrix pair is observable, there exists a non-singular matrix satisfying Define and A steady-state generator can be obtained as:

[0140]

[0141] wherein,

[0142] j(m) = diag(j1(m), j2(m),..., j N (m)), and Θ = diag(Θ1, Θ2,..., Θ N ).

[0143] (4) Based on the distributed internal model observer, design a distributed cooperative target encircling guidance law to guide the unmanned surface vehicle swarm system to achieve the target encircling mission. Define the relative velocity of the i-th unmanned surface vehicle and the target in the target body coordinate system as:

[0144]

[0145] wherein, u ri and v ri respectively represent the relative velocities of the i-th unmanned surface vehicle and the target in the longitudinal and lateral directions.

[0146] Define the encircling distance error ρ e_i and the encircling angle error η e_i as:

[0147] ρ e_i = ρ i - ρ d

[0148] θ e_i = η i - η d_i

[0149] The following distributed asymptotically average encircling guidance law is proposed:

[0150]

[0151] and

[0152]

[0153] wherein k 1_i 、k 2_i 、k p and Δ c represent positive design parameters, and u d represents the steady-state circumferential velocity. This guidance method realizes the collision-free circumvention of the target by the unmanned surface vehicle cluster system. This guidance method is applicable to periodic and aperiodic communication scenarios.

[0154] (5) Based on the distributed approximate average circumferential guidance law, a dynamic control method is designed for each unmanned surface vehicle under various constraints such as unknown model parameters, complex external disturbances, underactuated characteristics, and unknown nonlinear input saturation, so that the unmanned surface vehicle can track its desired state.

[0155] For the control input signal the smooth saturation model based on the dead zone operator is expressed as follows:

[0156]

[0157] wherein represents the positive density factor; represents the dead zone operator of the Prandtl-Ishlinskii model. p i (γ i ) represents the density function, where γ i represents the threshold parameter, satisfying and when γ i > R, the density is 0. Different density functions can represent different nonlinear characteristics, so as to meet the requirements of different nonlinear saturation constraints. Such as Figure 3 and 4 show the schematic diagrams of the input saturation constraints under different saturation models. It can be seen that the saturation model is divided into two parts. The first part is the linear invertible term, and the second part is the nonlinear hysteresis term.

[0158] Define the longitudinal velocity error of the i-th unmanned surface vehicle as the yaw angular velocity error as The dynamic error model of the i-th unmanned surface vehicle is:

[0159]

[0160] wherein

[0161] In practical engineering applications, the model parameter F i is an unknown term, which leads to the inability to design the control method. Therefore, by utilizing the online approximation characteristics of the radial basis neural network for any continuous function, the unknown model parameters and external disturbances can be approximated as:

[0162]

[0163] In the formula, W i * represents the ideal weight, H i (x) represents the output based on the Gaussian function, and ε i represents the estimation error.

[0164] Based on the radial basis neural network and combined with the minimum parameter learning method, an online learning algorithm for the unknown model is proposed, which is further expressed as:

[0165]

[0166] In the formula, ι i = ||W i *T ||, ψ i (x) = ||H i (x)||, D i = ||ε i ||.

[0167] Design the following control method:

[0168]

[0169] In the formula, c1 and c2 represent positive design parameters. Additionally, and respectively represent the estimated values of D i , ι i , p i (γ i ) and υ i . The adaptive control law is designed as follows:

[0170]

[0171] In the formula, κ1, κ2, γ1, γ2, c3 and c4 represent positive design parameters.

[0172] Technical effects

[0173] (1) The present invention proposes a distributed internal model observer to observe the unknown state of the target system. The introduction of the observer solves the problems brought by system complexity, partially unknown target state, and intermittent communication, and has greater practical engineering significance.

[0174] (2) The present invention proposes a distributed quasi-asymptotic average surrounding guidance law, which can guide the unmanned surface vehicle (USV) swarm system to surround a dynamic target based on the target observations. This method is a design scheme with continuously updated control inputs, enabling the USV swarm system to lock the target faster and more autonomously, improving the stability and efficiency of the system. In addition, the phase angle repulsion theory is introduced into the guidance law, which not only solves the problem of surrounding a variable-speed dynamic target but also avoids the collision problem among USVs, improving the safety of the system.

[0175] (3) The present invention proposes a model-free anti-disturbance control method under unknown nonlinear saturation input constraints, which ensures the finite-time stability of the USV dynamic system under multiple constraints. In the design of the control method, an input saturation constraint model based on the dead-zone operator model is proposed, and a control method is designed by combining the model based on the dead-zone operator with the dynamics of the USV system, solving the influence of unknown nonlinear input saturation constraints on the system stability. The control method proposed in the present invention does not require prior knowledge of the nonlinear characteristics of the input saturation constraints and is more flexible in practical applications.

[0176] The above further describes the technical solutions provided by the present invention through several specific embodiments to highlight the advantages and beneficial effects of the technical solutions provided by the present invention. However, the above several specific embodiments are not used as limitations on the present invention. Any reasonable modifications and improvements, combinations of embodiments, and equivalent replacements based on the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A distributed cooperative dynamic target surrounding control method for unmanned boats under multiple constraints, characterized in that Including: Step 1: Establish a motion model; Design a distributed internal model observer: In view of the intermittency and instability of the maritime communication environment, a distributed internal model observer is set for each unmanned boat, and the internal motion state of the target is estimated through the iterative update of the state data exchange between each other and the observation data of the boat itself. Step 2: Design a distributed asymptotically mean-like surrounding guidance algorithm: Perform distributed scheduling on the desired speed and desired heading of the unmanned boat, so that multiple unmanned boats can still maintain a safe surrounding formation when the speed or direction of the dynamic target changes. Step 3: Consider input saturation constraints and perform dynamic control with adaptive learning: Compensate for the input saturation of the unmanned boat's thruster or rudder, and perform real-time approximation and correction on external environmental interference and the uncertainty of the unmanned boat model, so that each unmanned boat is controlled according to the desired trajectory and motion speed. Step 4: Implement multi-constraint control: Complete the stable and collision-free surrounding of the dynamic target.

2. The distributed cooperative dynamic target surrounding control method for an unmanned surface vehicle under multiple constraints according to claim 1, wherein Specifically, Step 1 is: Obtain the position information, heading information and speed information of multiple unmanned boats of the target, as well as the position information and motion state information of the dynamic target, and uniformly represent the kinematic models of the unmanned boat and the target in the same coordinate system.

3. A method for distributed cooperative dynamic target surrounding control of an unmanned boat under multiple constraints according to claim 1, characterized in that, According to the target state information output by the distributed internal model observer, construct a surrounding distance, surrounding angle and phase angle repulsion strategy, and perform distributed scheduling on the desired speed and desired heading of the unmanned boat.

4. A method for distributed cooperative dynamic target surrounding control of an unmanned boat under multiple constraints according to claim 1, characterized in that Use the dead zone operator model to compensate for the input saturation of the unmanned boat's thruster or rudder, and at the same time combine the adaptive online learning algorithm to perform real-time approximation and correction on external environmental interference and the uncertainty of the unmanned boat model through neural network or minimum parameter learning methods, so that each unmanned boat is controlled according to the desired trajectory and motion speed given by the distributed asymptotically mean-like surrounding guidance algorithm.

5. A method for distributed cooperative dynamic target surrounding control of an unmanned surface vehicle under multiple constraints according to claim 1, characterized in that, Comprehensively execute the output result of the obtained distributed internal model observer, the desired state of the distributed asymptotically mean-like surrounding guidance algorithm, and the dynamic control process driven by adaptive learning to complete the stable and collision-free surrounding of the dynamic target.

6. An unmanned surface vehicle distributed cooperative dynamic target surrounding control device under multiple constraints, characterized in that, Including: Module 1: Establish a motion model; Design a distributed internal model observer: In view of the intermittency and instability of the maritime communication environment, a distributed internal model observer is set for each unmanned boat, and the internal motion state of the target is estimated through the iterative update of the state data exchange between each other and the observation data of the boat itself. Module 2: Design a distributed asymptotically mean-like surrounding guidance algorithm: Perform distributed scheduling on the desired speed and desired heading of the unmanned boat, so that multiple unmanned boats can still maintain a safe surrounding formation when the speed or direction of the dynamic target changes. Module 3: Consider input saturation constraints and perform dynamic control with adaptive learning: Compensate for the input saturation of the unmanned boat's thruster or rudder, and perform real-time approximation and correction on external environmental interference and the uncertainty of the unmanned boat model, so that each unmanned boat is controlled according to the desired trajectory and motion speed. Module 4: Implement multi-constraint control: Complete the stable and collision-free surrounding of the dynamic target.

7. A computer storage medium for storing a computing program, characterized in that, When the computer program is read by a computer, the computer executes the method described in claim 1.

8. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method described in claim 1.

9. A computer program product, as a computer program, characterized in that, When the computer program is executed, the method described in claim 1 is implemented.

Images