Self-adaptive switching control method for unmanned surface vehicle

Abstract

The self-adaptive switching control method for the unmanned surface vehicle comprises the following steps: establishing a mathematical model of the unmanned surface vehicle, and determining a control target of the unmanned surface vehicle; designing a virtual control law and a first Lyapunov alternative function for controlling the unmanned surface vehicle based on a backstepping method; designing a second Lyapunov alternative function, establishing an adaptive neural network controller model and a robust controller model, and designing control input of the unmanned surface vehicle switching control system; designing updating laws of an adaptive neural network controller and a robust controller, and designing a third Lyapunov function to perform stability analysis on the system; and testing and simulating the unmanned surface vehicle system in a simulation environment by using the adjusted control strategy. According to the control method, on the premise of ensuring the stability of the system, the path tracking error of the unmanned surface vehicle is reduced, and the energy consumption of the controller is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of control of unmanned surface boats, and in particular to an adaptive switching control method for unmanned surface boats. Background Art

[0002] In the field of smart navigation technology, AUV (Autonomous Underwater Vehicle) has broad application prospects in the fields of commerce, science and military, and can be used in marine operations such as marine safety, submarine rescue, submarine data collection, monitoring of marine pollution, surveying and mapping, marine fisheries, etc. Due to its small size, easy to control, good flexibility, high efficiency and low cost, easy to use, low requirements for the monitoring environment, autonomous navigation and obstacle avoidance, etc., AUV can replace humans to perform dangerous, time-consuming and heavy tasks.

[0003] Path tracking control is one of the key technologies of surface unmanned boats. It refers to tracking a planned route and can be used for distributed formation control. Its control goal is: starting from any given initial posture, the surface unmanned boat can finally stabilize and converge to a pre-planned desired path in space under the effective control of the controller. The control goal is only required in space and has nothing to do with time parameters. There is no strict requirement for the position on the path. However, the underwater environment where the surface unmanned boat is located is usually affected by factors such as ocean currents, underwater terrain and time-varying obstacles, and the nonlinearity and uncertainty of its motion model also pose challenges to the path tracking control algorithm, and even lead to the bad situation of unstable control. Therefore, in the case of uncertain models and unknown wind and wave disturbances, it is of great practical significance to realize the path tracking of surface unmanned boats that takes into account safety and economy. In order to solve the uncertainty and disturbance in the system, the existing methods mainly include PID control, fuzzy control, adaptive control, neural network control, anti-disturbance control, sliding mode control, model predictive control, etc. However, due to the nonlinearity, strong coupling and time-varying characteristics of surface unmanned boats, the existing path tracking control effect is not good.

[0004] A neural network is a collection of adjustable neurons or nodes. Due to its ability to process data, learn nonlinear systems and provide approximations, it has been widely used in path tracking of surface unmanned boats. On this basis, the adaptive neural network controller can handle nonlinear systems that do not meet the matching conditions, and relaxes the requirement that the unknown nonlinear function is linearly parameterized. The robust controller is a relatively mature theory that can provide strict guarantees for the stability and performance of the system. However, it relies heavily on the accuracy of the system model, and inaccurate models may lead to a decrease in controller performance. In various uncertain situations, in order to ensure the stability of the system, the design of the robust controller is often conservative. Therefore, the robust controller is suitable for systems that require strict stability and performance guarantees, especially when the system parameters and environment have a clear uncertainty range. The design is relatively conservative but has a solid theoretical foundation. At the same time, the adaptive neural network controller is suitable for complex nonlinear systems and environments with high uncertainty. It has adaptive and learning capabilities, but requires a large amount of data and computing resources, and the stability guarantee is relatively insufficient.

[0005] Most existing adaptive neural controllers can ensure the semi-globally consistent ultimate bounded stability (SGUUB) of the system under the condition that the neural network estimate is always valid. However, the conditions under which the neural network algorithm is always valid are difficult to verify in advance. If the initialization is inappropriate or subjected to external disturbances, the path tracking performance may deteriorate or even become unstable. A common solution is to use an additional robust controller outside the neural activity area to pull the system back to the transient state. However, these methods require that the adaptive neural network controller and the robust controller must always work together within the neural activity area, and there is inevitably the problem of control energy waste. On the other hand, the operation of the neural network is relatively complex, and other problems will be encountered in actual operation, such as the difficulty in determining the controller parameters, the large system computational load, the possibility of computational explosion problems, and the inability of the system state to converge within a limited time. Summary of the invention

[0006] In order to solve the above technical problems, the purpose of the present invention is to provide an adaptive switching control method for a surface unmanned boat. When the boat has limited output, unknown control input, and unknown dynamic model and disturbance, a switching control law is used to perform switching control between an adaptive neural network controller model and a robust controller model, thereby providing an efficient control scheme that takes into account both robustness and economy.

[0007] The present invention provides an adaptive switching control method for an unmanned surface boat, comprising:

[0008] Step 1: Establish a mathematical model of the surface unmanned boat and determine the control target of the surface unmanned boat;

[0009] Step 2: Design the virtual control law and the first Lyapunov candidate function for the control of the surface unmanned vehicle based on the backstepping method;

[0010] Step 3: Design the second Lyapunov candidate function, establish the adaptive neural network controller model and the robust controller model, and design the control input of the surface unmanned boat switching control system;

[0011] Step 4: Design the update law of the adaptive neural network controller and the robust controller, and design the third Lyapunov function to analyze the stability of the system;

[0012] Step 5: Use the adjusted control strategy to test and simulate the surface unmanned vehicle system in a simulated environment, collect system output, uncertain disturbances, internal dynamic information and input information, and finally analyze the control effect.

[0013] The adaptive switching control method of an unmanned surface boat of the present invention has the following beneficial effects:

[0014] The present invention defines a new Lyapunov function to ensure that it will not exceed the limit during the movement. At the same time, considering the rapid performance of trajectory error convergence, the output-limited control parameter is set to be time-varying, ensuring that the output-limited control parameter can be adjusted in time according to the real-time trajectory error. Considering the complexity of reducing the acquisition of uncertainty information of the internal dynamics and external disturbances of the surface unmanned boat model, an adaptive neural network is used to obtain the model information of the autonomous surface unmanned boat motion control, wherein only one neural network is used to compensate all unknown parts in each backstepping design process, thereby reducing the number of adaptive parameters in the controller, thereby simplifying the controller. Considering the effective conditions and energy consumption of the neural network compensation effect, an adaptive backstepping control method based on a switching control system is designed, which combines the model state to effectively switch autonomously between the robust controller and the neural network compensation method. Ultimately, the system has a smaller path tracking error, faster convergence speed and less energy consumption under the premise of ensuring stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flow chart of an adaptive switching control method for an unmanned surface boat. DETAILED DESCRIPTION

[0016] like Figure 1 As shown, an adaptive switching control method of an unmanned surface boat of the present invention comprises:

[0017] Step 1: Establish a mathematical model of the unmanned surface vehicle and determine the control objectives of the unmanned surface vehicle, specifically:

[0018] Step 1.1: Considering the uncertainty in the nonlinear dynamics model and the uncertainty in the ocean environment, the kinematic model of the three-degree-of-freedom motion of the surface unmanned vehicle is established as follows:

[0019]

[0020] in, It is expressed as the three-degree-of-freedom position vector of the surface unmanned vehicle in the inertial coordinate system; where x and y represent the position, and ψ represents the heading angle.

[0021]

[0022] Among them, J(ψ) is the rotation transformation matrix, which transforms the unmanned boat model from the hull coordinate system to the inertial coordinate system, and its dynamic relationship is:

[0023]

[0024] J T J=I

[0025] The antisymmetric matrix is ​​represented as:

[0026]

[0027] v c Indicates the speed of ocean currents in the inertial coordinate system; v: = [u, v, r] T It represents the speed of the unmanned surface vehicle relative to the fluid in the unmanned surface vehicle coordinate system, u is the forward speed of the unmanned surface vehicle at the current moment, v is the horizontal drift speed of the unmanned surface vehicle at the current moment, and r is the bow angular velocity of the unmanned surface vehicle at the current moment; they satisfy the relationship:

[0028] v=v g -v o

[0029] Among them, v g represents the actual speed of the boat in the surface unmanned boat coordinate system relative to the inertial coordinate system; v o Represents the ocean current velocity in the surface unmanned vehicle coordinate system relative to the inertial coordinate system.

[0030] Step 1.2: Considering the uncertainty in the nonlinear dynamic model and the uncertainty in the ocean environment, the dynamic model of the three-degree-of-freedom motion of the surface unmanned vehicle is established as follows:

[0031]

[0032] in, is the control input vector, τ u is the thrust, τ v is the swing force, τ r is the yaw moment; M = MT ∈R 3×3 is the mass inertia matrix of the system, is the skew-symmetric matrix of the Coriolis centripetal force: is the nonlinear damping matrix; represents the time-varying wind and wave disturbance, ω u represents the disturbance of the forward torque, ω v represents the disturbance of the lateral moment, ω r represents the disturbance of the steering torque; where the ocean disturbance has finite energy and satisfies |ω(t)|≤ω M ,so is a positive constant vector; Unmodeled dynamics.

[0033] Step 1.3: Determine the control target of the surface unmanned vehicle so that all signals in the surface unmanned vehicle system are consistent and bounded, ensuring that the path tracking error and the speed tracking error are arbitrarily small, specifically:

[0034] Define the desired parameterized path as η d (θ)=(x d (θ),y d (θ),ψ d (θ) T ∈R 3 is the desired parameterized path, θ∈R 3 is the path parameter, x d (θ) is the horizontal coordinate of the desired position of the surface unmanned vehicle, y d (θ) is the ordinate of the desired position of the unmanned surface vehicle, ψ d (θ) is the heading angle of the parametric path expected by the unmanned surface vehicle; its first-order derivative and second-order derivative with respect to θ are defined as:

[0035]

[0036]

[0037] Assume that η d (θ) is sufficiently smooth, and its second-order derivative with respect to θ is bounded, and satisfies the following conditions: That is, there exists a positive constant q1 such that the set holds:

[0038]

[0039] The control goal is to design the control law so that all signals in the closed-loop system are consistent and ultimately bounded, and by selecting appropriate design parameters, the path tracking error and the speed tracking error can be made arbitrarily small, that is:

[0040] a)

[0041] b)

[0042] Among them, v d is the expected speed of the system, μ1,μ2∈R are positive constants.

[0043] Step 2: Design the virtual control law and the first Lyapunov candidate function for the control of the surface unmanned vehicle based on the backstepping method, specifically:

[0044] Step 2.1: Design virtual control law α1:

[0045]

[0046] Among them, the positive definite matrix Its diagonal elements are positive constants, that is,

[0047] Define path tracking error: z1 = J T (η-η d );

[0048] Define velocity tracking error:

[0049] Define the ocean current velocity estimation error: is an estimate of the ocean current speed;

[0050] Then take the derivative of z1 and we get:

[0051]

[0052] Step 2.2: Define the first Lyapunov candidate function:

[0053]

[0054] Taking the derivative of V1, we get:

[0055]

[0056] Define the second error: z2 = v-α1, replace the speed variable v with the virtual control law α1 and the second error z2, then It is expressed as:

[0057]

[0058] For all z1, but It is expressed as:

[0059]

[0060] Step 3: Design the second Lyapunov candidate function, establish the adaptive neural network controller model and the robust controller model, and design the control input of the surface unmanned boat switching control system, specifically:

[0061] Step 3.1: Design the following second Lyapunov candidate function:

[0062]

[0063] Among them, the weight matrix estimation error of the adaptive neural network controller is Indicates that the neural network estimator has l nodes; the symmetric matrix In the subsequent design of the control input, the estimated errors of the two adjustable parameters are integrated into in represents the estimate of the adjustable parameter in the control input of the subsequent design, where the two-dimensional vector is the optimal parameter vector of adjustable parameters; accordingly, is the estimate of the tuning parameters of the adaptive neural network part, where for The theoretical optimal value of for The parameter estimation error of is the estimation of the tuning parameters of the robust control part, where for The theoretical optimal value of for Parameter estimation error; designed positive parameter δ1;

[0064] Derivative of V2 and parallel have to:

[0065]

[0066] in, Represents the weight matrix estimate of the adaptive neural network controller.

[0067] Considering the time derivative of z2, we get:

[0068]

[0069] in, It consists of three parts:

[0070]

[0071] Define function h1(P1) containing The more difficult part to obtain accurately is:

[0072]

[0073] Defining variables

[0074] According to the assumption: for the smooth function g(x i ) There exists a known positive smooth function and a positive constant g, then for

[0075]

[0076] In addition, regarding g(x i ) for The following inequalities are satisfied:

[0077]

[0078] Among them, the unknown constant ∈ i * ≥0.

[0079] Thus, there exists a positive smooth function γ1(P1) and an unknown constant The following relations are satisfied:

[0080]

[0081] Thus, a radial basis neural network is used to estimate h1(P1):

[0082]

[0083] Among them, variable P1 is the input vector of this radial basis neural network, represents the ideal constant weight matrix, l>1 is the number of nodes in the neural network; S1(P1)=[s1(P1),…,s l (P1)] T :Ω i →R l , S1(P1) is a known smooth vector function, s q (P1), q=1,...,l is a commonly used Gaussian function, expressed as:

[0084]

[0085] μ q is the center of the neural network receptive field, η q >0 is the dispersion of Gaussian function;

[0086] Ideal optimal constant weight matrix for The estimated value of is as follows:

[0087]

[0088] The estimated error ∈1(P1) satisfies is an unknown constant greater than 0; the adaptive neural network controller designed subsequently is simplified based on this radial basis neural network.

[0089] Step 3.2: Design the second Lyapunov candidate function, establish the adaptive neural network controller model and the robust controller model, and design the control input of the surface unmanned boat switching control system:

[0090]

[0091] The adaptive neural network controller is:

[0092]

[0093] The robust controller is:

[0094]

[0095] in, k2>0.

[0096] The robust controller takes over the control outside the effective area estimated by the neural network controller, thereby ensuring the stability of the system and improving the performance of the system. The switching between the two controllers is achieved through the switching function, which is as follows:

[0097]

[0098] Among them, r2>r1>0.

[0099] Substitute the obtained control input of the switching control system into have to:

[0100]

[0101]

[0102] According to the property of the switching function 0≤m(P1)≤1, the second Lyapunov candidate function inequality is obtained:

[0103]

[0104]

[0105] Among them design

[0106] Step 4: Design the update law of the adaptive neural network controller and the robust controller, and design the third Lyapunov function to analyze the stability of the system, specifically:

[0107] Step 4.1: Design update law:

[0108]

[0109] Among them, σ1 is the design parameter.

[0110] Substituting the update law into the second Lyapunov candidate function inequality, we get:

[0111]

[0112] According to Schwartz inequality:

[0113]

[0114] Among them, κ=0.2785, which is a constant, satisfies κ=e -(κ+1) .

[0115] We get the inequality:

[0116]

[0117] in:

[0118]

[0119] Considering e v Influenced by, redesign the third Lyapunov function:

[0120]

[0121] Taking the derivative of the third Lyapunov function, we get:

[0122]

[0123] Choose appropriate parameters:

[0124]

[0125] thus:

[0126]

[0127] From this we can see that the derivative of the system's Lyapunov function is bounded, and the system's Lyapunov stability is proven.

[0128] Step 5: Use the adjusted control strategy to test and simulate the surface unmanned vehicle system in a simulated environment, collect system output, uncertain disturbances, internal dynamic information and input information, and finally analyze the control effect.

[0129] On the basis of traditional backstepping method and Lyapunov stability, the present invention establishes a nonlinear dynamic model of the surface unmanned boat, uses an adaptive neural network controller to estimate and compensate for the uncertain parts of the model such as the uncertain parameters and modeling errors of the surface unmanned boat, and designs a switching control system based on an adaptive neural network controller and a robust controller to improve system performance. Under the premise of ensuring system stability, the error of the path tracking of the surface unmanned boat is reduced and the energy consumption of the controller is reduced.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An adaptive switching control method for an unmanned surface vehicle, characterized in that: include: Step 1: Establish a mathematical model of the surface unmanned boat and determine the control target of the surface unmanned boat; Step 2: Design the virtual control law and the first Lyapunov candidate function for the control of the surface unmanned vehicle based on the backstepping method; Step 3: Design the second Lyapunov candidate function, establish the adaptive neural network controller model and the robust controller model, and design the control input of the surface unmanned boat switching control system; Step 4: Design the update law of the adaptive neural network controller and the robust controller, and design the third Lyapunov function to analyze the stability of the system; Step 5: Use the adjusted control strategy to test and simulate the surface unmanned vehicle system in a simulated environment, collect system output, uncertain disturbances, internal dynamic information and input information, and finally analyze the control effect.

2. The adaptive switching control method of the unmanned surface vehicle according to claim 1, characterized in that: The step 1 is specifically as follows: Step 1.1: Considering the uncertainty in the nonlinear dynamics model and the uncertainty in the ocean environment, the kinematic model of the three-degree-of-freedom motion of the surface unmanned vehicle is established as follows: in, It is expressed as the position vector of the three degrees of freedom of the surface unmanned vehicle in the inertial coordinate system; where x, y represent the position, and ψ represents the heading angle; Among them, J(ψ) is the rotation transformation matrix, which transforms the unmanned boat model from the hull coordinate system to the inertial coordinate system, and its dynamic relationship is: J T J=I The antisymmetric matrix is ​​represented as: v c Indicates the speed of ocean currents in the inertial coordinate system; v: = [u, v, r] T It represents the speed of the unmanned surface vehicle relative to the fluid in the unmanned surface vehicle coordinate system, u is the forward speed of the unmanned surface vehicle at the current moment, v is the horizontal drift speed of the unmanned surface vehicle at the current moment, and r is the bow angular velocity of the unmanned surface vehicle at the current moment; they satisfy the relationship: v=v g -v o Among them, v g represents the actual speed of the boat in the surface unmanned boat coordinate system relative to the inertial coordinate system; v o It represents the ocean current velocity in the surface unmanned vehicle coordinate system relative to the inertial coordinate system; Step 1.2: Considering the uncertainty in the nonlinear dynamic model and the uncertainty in the ocean environment, the dynamic model of the three-degree-of-freedom motion of the surface unmanned vehicle is established as follows: in, is the control input vector, τ u is the thrust, τ v is the swing force, τ r is the yaw moment; M = M T ∈R 3×3 is the mass inertia matrix of the system, is the skew-symmetric matrix of the Coriolis centripetal force: is the nonlinear damping matrix; represents the time-varying wind and wave disturbance, ω u represents the disturbance of the forward torque, ω v represents the disturbance of the lateral moment, ω r represents the disturbance of the steering torque; where the ocean disturbance has finite energy and satisfies |ω(t)|≤ω M ,so is a positive constant vector; for unmodeled dynamics; Step 1.3: Determine the control target of the surface unmanned vehicle so that all signals in the surface unmanned vehicle system are consistent and bounded, ensuring that the path tracking error and the speed tracking error are arbitrarily small, specifically: Define the desired parameterized path as η d (θ)=(x d (θ),y d (θ),ψ d (θ) T ∈R 3 is the desired parameterized path, θ∈R 3 is the path parameter, x d (θ) is the horizontal coordinate of the desired position of the surface unmanned vehicle, y d (θ) is the ordinate of the desired position of the unmanned surface vehicle, ψ d (θ) is the heading angle of the parametric path expected by the unmanned surface vehicle; its first-order derivative and second-order derivative with respect to θ are defined as: Assume that η d (θ) is sufficiently smooth, and its second-order derivative with respect to θ is bounded, and satisfies the following conditions:, That is, there exists a positive constant v1 such that the set holds: The control goal is to design the control law so that all signals in the closed-loop system are consistent and ultimately bounded, and by selecting appropriate design parameters, the path tracking error and the speed tracking error can be made arbitrarily small, that is: a) b) Among them, v d is the expected speed of the system, μ1,μ2∈T are positive constants.

3. The adaptive switching control method of the unmanned surface vehicle according to claim 2, characterized in that: The step 2 is specifically as follows: Step 2.1: Design virtual control law α1: Among them, the positive definite matrix Its diagonal elements are positive constants, that is, Define path tracking error: z1 = J T (η-η d ); Define velocity tracking error: Define the ocean current velocity estimation error: is an estimate of the ocean current speed; Then take the derivative of z1 and we get: Step 2.2: Define the first Lyapunov candidate function: Taking the derivative of V1, we get: Define the second error: z2 = v-α1, replace the speed variable v with the virtual control law α1 and the second error z2, then It is expressed as: For all z1, but It is expressed as:

4. The adaptive switching control method of the unmanned surface vehicle according to claim 3, characterized in that: The step 3 is specifically as follows: Step 3.1: Design the following second Lyapunov candidate function: Among them, the weight matrix estimation error of the adaptive neural network controller is Indicates that the neural network estimator has l nodes; the symmetric matrix In the subsequent design of the control input, the estimated errors of the two adjustable parameters are integrated into in represents the estimate of the adjustable parameter in the control input of the subsequent design, where the two-dimensional vector is the optimal parameter vector of adjustable parameters; accordingly, is the estimate of the tuning parameters of the adaptive neural network part, where for The theoretical optimal value of for The parameter estimation error of is the estimation of the tuning parameters of the robust control part, where for The theoretical optimal value of for Parameter estimation error; designed positive parameter δ1; Derivative of V2 and parallel have to: in, Represents the weight matrix estimate of the adaptive neural network controller; Considering the time derivative of z2, we get: in, It consists of three parts: Define function h1(P1) containing The more difficult part to obtain accurately is: Defining variables According to the assumption: for the smooth function g(x i ) There exists a known positive smooth function and the positive constant g , then for In addition, regarding g(x i ) for The following inequalities are satisfied: Among them, the unknown constant ∈ i * ≥0; Thus, there exists a positive smooth function γ1(P1) and an unknown constant The following relations are satisfied: Thus, a radial basis neural network is used to estimate h1(P1): Among them, variable P1 is the input vector of this radial basis neural network, represents the ideal constant weight matrix, l>1 is the number of nodes in the neural network; S1(P1)=[s1(P1),…,s l (P1)] T :Ω i →R l , S1(P1) is a known smooth vector function, s q (P1), q=1,...,l is a commonly used Gaussian function, expressed as: μ q is the center of the neural network receptive field, η q >0 is the dispersion of Gaussian function; Ideal optimal constant weight matrix for The estimated value of is as follows: The estimated error ∈1(P1) satisfies is an unknown constant greater than 0; the adaptive neural network controller designed subsequently is simplified based on this radial basis neural network; Step 3.2: Design the second Lyapunov candidate function, establish the adaptive neural network controller model and the robust controller model, and design the control input of the surface unmanned boat switching control system: The adaptive neural network controller is: The robust controller is: in, The robust controller takes over the control outside the effective area estimated by the neural network controller, thereby ensuring the stability of the system and improving the performance of the system. The switching between the two controllers is achieved through the switching function, which is as follows: Among them, r2>r1>0; Substitute the obtained control input of the switching control system into have to: According to the property of the switching function 0≤m(P1)≤1, the second Lyapunov candidate function inequality is obtained: Among them design 5. The adaptive switching control method of the unmanned surface vehicle according to claim 4, characterized in that: The step 4 is specifically as follows: Step 4.1: Design the update law for the adaptive neural network controller and the robust controller: Among them, σ1 is the design parameter; Substituting the update law into the second Lyapunov candidate function inequality, we get: According to Schwartz inequality and Polycarpou's inequality: Among them, κ=0.2785, which is a constant, satisfies κ=e -(κ+1) ; We get the inequality: in: Considering e v Influenced by, redesign the third Lyapunov function: Taking the derivative of the third Lyapunov function, we get: Choose appropriate parameters and design a positive parameter K ω : thus: From this we can see that the derivative of the system's Lyapunov function is bounded, and the system's Lyapunov stability is proven.

Images