A ship dynamic positioning control system and method based on finite time command filtering backstepping method

Abstract

The present invention provides a ship dynamic positioning control system and method based on a finite time instruction filter backstepping method, comprising a display and control computer (1), a finite time instruction filter backstepping controller (2), an RBF neural network error compensation system (3), a propulsion system (4), a dynamic positioning ship (5), and a sensor system (6). The ship dynamic positioning control system and method based on a finite time instruction filter backstepping method of the present invention firstly considers the unmodeled dynamics and errors in the ship motion and the dynamic characteristics of the propeller, and transforms the control system into a third-order system; secondly, an adaptive neural network method is used to estimate the unknown interference in the system; finally, a finite time control technology is introduced in the design of the instruction filter backstepping controller, so that it has a finite time characteristic, and Lyapunov stability theory is used to prove that the system tracking error and parameter estimation error can converge within a finite time.

Description

Technical Field

[0001] The present invention relates to a finite time instruction filtering backstepping control system for ship dynamic positioning, and also relates to a finite time instruction filtering backstepping control method for ship dynamic positioning. Background Art

[0002] With the continuous development of deep-sea resources, research on ship positioning methods has become increasingly important. Dynamic positioning systems are a key area of ​​research. The control system is a crucial component of dynamic positioning systems, and therefore, nonlinear dynamic positioning (DP) control has attracted considerable attention. Common DP control methods include PID control, LQR control, model predictive control, and robust adaptive control. Robust adaptive control has been a recent hot topic of research, with key adaptive control methods including sliding mode control and adaptive backstepping.

[0003] Chinese patent CN109460043A proposes a method for ship trajectory self-disturbance rejection control based on a multi-modal non-singular terminal sliding mode. This patent utilizes a state observer to enable its internal function to switch between linear and nonlinear functions in real time; a switchable segmented sliding surface is designed, and a new bi-power convergence law is introduced to achieve the purpose of rapid convergence. Finally, a switchable error feedback control law is designed using this segmented sliding surface. The sliding mode control method has good robustness and fast convergence, but it can only handle specific systems and is not universal. Unlike this method, the present invention mainly uses a finite-time command filtering method to study the DP system based on model uncertainty, unmodeled dynamic terms, and propeller dynamic characteristics in the ship motion under the actuator. Summary of the Invention

[0004] The present invention aims to provide a neural network-based finite-time instruction filtering backstepping positioning control system with good anti-interference capability, high positioning accuracy, and the ability to quickly issue instructions to an actuator within a limited time to achieve a good control effect. The present invention also aims to provide a neural network-based finite-time instruction filtering backstepping positioning control method.

[0005] The present invention discloses a ship dynamic positioning control system based on a finite time instruction filter backstepping method, comprising a display and control computer (1), a finite time instruction filter backstepping controller (2), an RBF neural network error compensation system (3), a propulsion system (4), a dynamic positioning ship (5), and a sensor system (6). The present invention is characterized in that: the sensor system (6) collects position information η of the dynamic positioning ship (5) in real time, and transmits the sampled position information to the display and control computer (1), the finite time instruction filter backstepping controller (2), and the RBF neural network compensation system (3); the display and control computer (1) displays the actual position information of the ship in real time and transmits the expected position step signal μ to the display and control computer (1). d is passed to the finite time instruction filter backstepping controller (2); the finite time instruction filter backstepping controller (2) constructs the virtual signal β i and virtual control quantity μ c , for the desired position step signal η d Calculate and obtain each tracking error η e 、 μ e , construct error compensation signals ξ1, ξ2, ξ3 and pass them to each tracking error compensation vector system to output each tracking error compensation vector v1, v2, v3; RBF neural network compensation system (3) approximates the nonlinear function F(η, θ), and performs RBF approximation on the control input error θ to obtain the adaptive update rate and passed to the finite time command filter backstepping controller (2); adaptive update rate and the virtual signal β i Finite time filtering control is performed to obtain the control instruction τ of the propulsion system (4) under the dynamic characteristics of the propeller c The propulsion system (4) outputs the control command τ according to the finite time command filter backstepping controller (2) c The dynamic positioning vessel (5) is controlled and the RBF neural network error compensation system (3) is fed back so that the vessel moves to a desired position.

[0006] The ship dynamic positioning control method based on the finite time instruction filtering backstepping method of the present invention is:

[0007] (1) The sensor system (6) collects the position information η of the dynamically positioned vessel (5) in real time, and transmits the sampled position information to the display and control computer (1), the finite time instruction filter backstepping controller (2) and the RBF neural network compensation system (3);

[0008] (2) The display and control computer (1) displays the actual position information of the ship in real time and transmits the desired position step signal η d Passed to the finite time command filter backstepping controller (2);

[0009] (3) Finite time command filter backstepping controller (2) Construct virtual signal β i and virtual control quantity μ c For the desired position step signal η d Calculate and obtain each tracking error η e 、 μ e , construct error compensation signals ξ1, ξ2, ξ3 and transmit them to each tracking error compensation vector system to output each tracking error compensation vector ν1, v2, v3;

[0010] (4) RBF neural network compensation system (3) approximates the nonlinear function F(η, θ), performs RBF approximation on the control input error θ, and obtains the adaptive update rate And pass it to the finite time command filter backstepping controller (2);

[0011] (5) Adaptive update rate and the virtual signal β i Finite time filtering control is performed to obtain the control instruction τ of the propulsion system (4) under the dynamic characteristics of the propeller c ;

[0012] (6) The propulsion system (4) outputs the control command τ according to the finite time command filtering backstepping controller (2) c The dynamic positioning vessel (5) is controlled and the RBF neural network error compensation system (3) is fed back so that the vessel moves to a desired position.

[0013] The present invention can improve the anti-interference capability and positioning accuracy of a ship's dynamic positioning system, and can achieve better control effects under the conditions of considering model uncertainty, unmodeled dynamic terms and dynamic characteristics of the propeller. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is the overall structure diagram of the ship dynamic positioning control system based on the finite time command filter backstepping method. DETAILED DESCRIPTION

[0015] The present invention discloses a ship dynamic positioning control system based on a finite time instruction filter backstepping method, comprising a display and control computer (1), a finite time instruction filter backstepping controller (2), an RBF neural network error compensation system (3), a propulsion system (4), a dynamic positioning ship (5), and a sensor system (6). The present invention is characterized in that: the sensor system (6) collects position information η of the dynamic positioning ship (5) in real time, and transmits the sampled position information to the display and control computer (1), the finite time instruction filter backstepping controller (2), and the RBF neural network compensation system (3); the display and control computer (1) displays the actual position information of the ship in real time and transmits the expected position step signal η to the display and control computer (1), the finite time instruction filter backstepping controller (2), and the RBF neural network compensation system (3); the display and control computer (1) displays the actual position information of the ship in real time and transmits the expected position step signal η to the display and control computer (1). d is passed to the finite time instruction filter backstepping controller (2); the finite time instruction filter backstepping controller (2) constructs the virtual signal β i and virtual control quantity μ c , for the desired position step signal η d Calculate and obtain each tracking error η e 、 μ e , construct error compensation signals ξ1, ξ2, ξ3 and pass them to each tracking error compensation vector system to output each tracking error compensation vector v1, v2, ν3; RBF neural network compensation system (3) approximates the nonlinear function F(η, θ), and performs RBF approximation on the control input error θ to obtain the adaptive update rate and passed to the finite time command filter backstepping controller (2); adaptive update rate and the virtual signal β i Finite time filtering control is performed to obtain the control instruction τ of the propulsion system (4) under the dynamic characteristics of the propeller c The propulsion system (4) outputs the control command τ according to the finite time command filter backstepping controller (2) c The dynamic positioning vessel (5) is controlled and the RBF neural network error compensation system (3) is fed back so that the vessel moves to a desired position.

[0016] The following is a more detailed description of the ship dynamic positioning control method based on the finite time instruction filter backstepping method of the present invention:

[0017] The kinematic model and dynamic model of each following ship are established separately.

[0018]

[0019] where η = [xyψ] T is the position and heading of the ship in the geodetic coordinate system; υ = [uvr] T is the velocity vector of the ship in the hull coordinate system; R is the rotation matrix and M is the inertial mass matrix; C(u) is the Coriolis and centripetal force matrix; D(υ) is the damping matrix; τ is the force and torque generated by the propulsion system; d(t) represents the unknown disturbance, which includes ocean disturbances caused by wind, waves, and currents; ω represents the unmodeled dynamics and errors of the system and

[0020] In controller design, it is necessary to consider the inertial characteristics of the thruster to eliminate the influence of control command lag on the system. We can consider the thruster as a first-order inertial link, and its model can be expressed as:

[0021]

[0022] Among them: A and B are diagonal matrices; τ c is the control instruction. In order to facilitate the controller design, we first need to transform formulas (1) and (2).

[0023] make

[0024]

[0025] and Then we can get:

[0026]

[0027] in

[0028] Define each tracking error as

[0029]

[0030] where η d It is the desired position and heading set in advance; u c =[u c1 ,u c2 ,u c3 ] T The virtual control quantity is generated by the following finite time command filter. This filter can approximate the differential signal of the virtual control quantity faster than the traditional filter.

[0031]

[0032] where β i is the filter input, which will be designed later.

[0033] The compensation vector of each tracking error is defined as:

[0034]

[0035] where ξ1,ξ2,ξ3 are error compensation signals and are defined as

[0036]

[0037] where b i ,c i ,γ is a positive real number and 0<γ<1,β i (i=1,2,3) is a virtual signal and is constructed as

[0038]

[0039] where sig γ (v i )=[|v i1 | γ sgn(v i1 ),|v i2 | γ sgn(v i2 ),|v i3 | γ sgn(v i3 )] T ,i=1,2,3

[0040] Introducing an adaptive neural network system to Estimate the unknown model parameters in .

[0041] Using RBF neural network to analyze nonlinear functions Approximating this, the neural network output can be expressed as:

[0042] F i (X) = W fi t Φ fi (X)(9)

[0043] According to the approximation principle, for a compact set For any X on , the output of RBF can approximate F(X) with arbitrary precision, that is:

[0044]

[0045] in is the ideal weight vector

[0046]

[0047] ε fi is the approximation error, and there exists a positive constant δ f , making And for j = 1, 2, ..., n, is a Gaussian function, e j =[c j1 ,c j2 ,…,c jm ] T is the center vector of the Gaussian function of the i-th neuron, d i is the basis width of the Gaussian function.

[0048] So β2 can be rewritten as

[0049]

[0050] in

[0051] Define θ=max{||W f1 || 2 ,||W f2 || 2 ,||W f3 || 2}, We can choose The adaptive update rate is

[0052]

[0053] For systems (1) and (2), we can choose the finite-time command filter (5), the error compensation system (7), the virtual control signals (8) and (11), the control law in (8) and the adaptive update rate (12) so that the tracking error in (4), the tracking error compensation vector in (6), and the estimation error of the adaptive neural network system for the nonlinear function are Finite time convergence.

[0054] According to (4), the derivative of the tracking error can be obtained

[0055]

[0056] Taking the derivative of the compensation system and substituting (7) and (8) into it, we can obtain

[0057]

[0058] We choose the following Lyapunov function:

[0059]

[0060] Consider (14) and take its derivative

[0061]

[0062] According to Young's inequality, we can get the following inequality

[0063]

[0064] At the same time, according to the theorem, we can get the following inequality

[0065]

[0066] Substituting (12), (17), (18), (19) and (20) into (16), we obtain

[0067]

[0068] The following two cases are discussed:

[0069] (1) If We can get the following inequality

[0070]

[0071] (2) If Then we can get the inequality:

[0072]

[0073] Therefore, based on the above two points, we can get

[0074]

[0075] in

[0076] and

[0077] So we can get the following inequality:

[0078]

[0079] From Lemma 1, we can get the following conclusion: The compensation errors v1, v2, v3 are Convergence within the region From the definition of formula (10), we know that if we prove ξ i is bounded in finite time, then the tracking error in (6) can converge to the vicinity of the origin in a finite time.

[0080] Next, we prove that the error compensation system is bounded in finite time. First, we choose the following Lyapunov function

[0081]

[0082] Taking the derivative of it, we can get

[0083]

[0084] According to the lemma, we know that in finite time So we have

[0085]

[0086] Where k0=min(2b i ), As long as l0 is chosen appropriately so that Then ξ i is finite-time stable.

Claims

1. A ship dynamic positioning control system based on a finite time instruction filter backstepping method, comprising a display and control computer (1), a finite time instruction filter backstepping controller (2), an RBF neural network error compensation system (3), a propulsion system (4), a dynamic positioning ship (5), and a sensor system (6), wherein: The sensor system (6) collects the position and posture information η of the dynamically positioned ship (5) in real time, and transmits the sampled position and posture information to the display and control computer (1), the finite time instruction filter backstepping controller (2) and the RBF neural network compensation system (3); the display and control computer (1) displays the actual position and posture information of the ship in real time and transmits the expected position step signal η to the display and control computer (1). d is passed to the finite time instruction filter backstepping controller (2); the finite time instruction filter backstepping controller (2) constructs a virtual signal β i and virtual control volume μ c , for the expected position step signal η d Calculate and obtain each tracking error η e , μ e , construct error compensation signals ξ1, ξ2, ξ3 and pass them to each tracking error compensation vector system to output each tracking error compensation vector ν1, ν2, ν3; RBF neural network compensation system (3) approximates the nonlinear function F(η, θ), and performs RBF approximation on the control input error θ to obtain the adaptive update rate and passed to the finite time command filter backstepping controller (2); adaptive update rate and the virtual signal β i Finite time filtering control is performed to obtain the control command τ of the propulsion system (4) under the dynamic characteristics of the propeller c The propulsion system (4) outputs the control command τ according to the finite time command filter backstepping controller (2). c The dynamic positioning ship (5) is controlled and the RBF neural network error compensation system (3) is fed back so that the ship moves to a desired position.

2. A control method for a ship dynamic positioning control system based on a finite time instruction filter backstepping method according to claim 1, characterized in that: (1) The sensor system (6) collects the position information η of the dynamic positioning vessel (5) in real time, and transmits the sampled position information to the display control computer (1), the finite time instruction filter backstepping controller (2) and the RBF neural network compensation system (3); (2) The display and control computer (1) displays the actual position information of the ship in real time and converts the expected position step signal η d Passed to the finite time command filter backstepping controller (2); (3) Finite time command filter backstepping controller (2) Construct virtual signal β i and virtual control volume μ c , for the expected position step signal η d Calculate and obtain each tracking error η e , μ e , construct error compensation signals ξ1, ξ2, ξ3 and transmit them to each tracking error compensation vector system to output each tracking error compensation vector v1, v2, v3; (4) RBF neural network compensation system (3) Approximate the nonlinear function F(μ, θ), perform RBF approximation on the control input error θ, and obtain the adaptive update rate And passed to the finite time command filter backstepping controller (2); (5) Adaptive update rate and the virtual signal β i Finite time filtering control is performed to obtain the control command τ of the propulsion system (4) under the dynamic characteristics of the propeller c ; (6) The propulsion system (4) outputs the control command τ according to the finite time command filtering backstepping controller (2) c The dynamic positioning ship (5) is controlled and the RBF neural network error compensation system (3) is fed back so that the ship moves to a desired position.

Images