Self-adaptive neural network course tracking control method for unmanned surface vehicle

Abstract

The invention provides a self-adaptive neural network course tracking control method for an unmanned surface vehicle, and the method comprises the steps: building a third-order response type unmanned surface vehicle course control mathematical model according to the surrounding environment and the sea condition information of other surrounding ships; a linear input quantitative analysis model is introduced to quantify control input in a control system; based on a dynamic surface control algorithm, designing a Lyapunov function, and obtaining a course controller of the unmanned surface vehicle; based on the Lyapunov stability theory, the stability of the designed adaptive neural network course tracking control system of the unmanned surface vehicle with input quantization is proved without fixed priori information of quantization parameters, and all signals in a closed-loop system are consistent and bounded in a semi-global manner. And the tracking error can be arbitrarily small by adjusting the parameters of the controller. According to the technical scheme, the course control performance of the unmanned surface vehicle can be improved, and higher universality and adaptability are achieved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an adaptive neural network heading tracking control method for an unmanned surface vehicle. Background Art

[0002] With the rise of smart maritime shipping, unmanned surface vehicles (USVs) have become an important part of global maritime operations. Their strong maneuverability, advanced autonomous navigation capabilities, ability to carry multifunctional payloads, and ability to perform monitoring, rescue, and remote dangerous tasks in vast seas have made them a key player in maritime transportation missions.

[0003] In a complex and changing marine environment, maintaining the stability of the ship and accurately controlling its heading is one of the main challenges for unmanned surface vehicles to achieve autonomous navigation. At present, most research is committed to improving the tracking stability and accuracy of unmanned surface vehicles by designing efficient and applicable control strategies. However, since the various components of unmanned surface vehicles rely on communication channels to transmit information, and the bandwidth of communication channels is limited, they face the problem of insufficient communication resources. By introducing input quantization technology before signal transmission, the operating frequency of the actuator can be effectively reduced, and the communication rate within the signal transmission bandwidth can be reduced. Therefore, in the research on heading control of unmanned surface vehicles, the application of input quantization technology is gradually receiving widespread attention.

[0004] However, simply relying on input quantization technology for heading control cannot fully meet the development needs of unmanned surface vehicles. On the one hand, since the ship's steering system cannot achieve step steering, ignoring the steering characteristics will affect the performance quality of the control system. Therefore, from a practical application point of view, in order to obtain good heading control performance, the characteristics of the steering servo system should also be considered. On the other hand, after introducing the linear input quantization analysis model, the dynamic surface control (DSC) algorithm is used to solve the problem of "differential explosion" in the traditional inversion method. In addition, effective smoothing of the quantized signal can also be achieved. Summary of the invention

[0005] According to the above-mentioned technical problems such as the limited communication bandwidth of unmanned surface vehicles at sea and the prevention of collisions caused by position errors, an unmanned surface vehicle adaptive neural network heading tracking control method with actuator input quantization and ship steering gear dynamic characteristics is provided, aiming to solve the model uncertainty in the control system and the external interference problems in heading control. The present invention mainly uses a linear model to describe the quantization process, which effectively reduces the execution frequency of the actuator, thereby saving communication resources and reducing communication pressure. In addition, a system adaptive control law is designed to successfully minimize the tracking error and ensure that the unmanned surface vehicle can accurately navigate along the predetermined route; and using Lyapunov stability theory, it is proved that all signals of the closed-loop system are semi-globally uniformly bounded.

[0006] The technical means adopted by the present invention are as follows:

[0007] An adaptive neural network heading tracking control method for an unmanned surface vehicle, comprising:

[0008] S1. According to the surrounding environment and the sea conditions of other ships around, a third-order response unmanned surface vehicle heading control mathematical model is established;

[0009] S2, introduce the linear input quantification analysis model to quantify the control input in the control system;

[0010] S3. Based on the dynamic surface control algorithm, design the Lyapunov function to obtain the heading controller of the unmanned surface vehicle;

[0011] S4. Based on Lyapunov stability theory, the stability of the designed unmanned surface vehicle adaptive neural network heading tracking control system with input quantization is proved when there is no need for fixed prior information of quantization parameters, and all signals in the closed-loop system are semi-globally uniformly bounded. The tracking error can be made arbitrarily small by adjusting the controller parameters.

[0012] Furthermore, step S1 specifically includes:

[0013] S11. Construct a ship nonlinear Nomoto model and a ship steering gear dynamic characteristics model as follows:

[0014]

[0015] Among them, ψ represents the ship's heading angle, δ is the rudder angle, K and T represent the ship's motion maneuverability index, α and β represent nonlinear parameters, and δ E Indicates the command rudder angle issued by the servo, K E Indicates the servo control gain, T E Indicates the servo time constant;

[0016] S12. Introduce input quantization and construct a third-order response mathematical model of the unmanned surface vehicle as follows:

[0017]

[0018] in, ω1 and ω2 both represent unknown disturbances caused by wind, current and waves.

[0019] Furthermore, step S2 specifically includes:

[0020] S21. For the control input after considering the input quantization in the ship model, let Q(u) = q1(t)u + q2(t), and take:

[0021]

[0022] S22. Since the sign of the quantization process remains unchanged, q1(t) is unknown. From the formula in step S21, we know that q1(t)>0; when |u(t)|<a, Q(u) is bounded, q1(t)=1, then q2(t) is bounded, so we take

[0023] Furthermore, step S3 specifically includes:

[0024] S31, define heading error as: S1 = x1-x ld , S2=x2-x 2d , S3=x3-x 3d , and respectively differentiate the heading error to obtain:

[0025]

[0026] S32. Define the Lyapunov function as follows:

[0027]

[0028] Among them, y i represents the error of the virtual control item,

[0029] S33. Define a first-order low-pass filter as follows:

[0030]

[0031] in,

[0032] S34. Since f(x,t) is unknown, a RBF neural network is constructed to approximate the unknown function of the system model. There exists an ideal weight vector θ * ∈R N , so that the neural network θ *Th(x) is close enough to the given function f and the absolute value of the approximation error is no greater than σ M ,Right now:

[0033] f(x)=θ *T h(x)+ε *

[0034] S35. According to Young's inequality, the derivatives of V1 and V2 in step S32 are obtained:

[0035]

[0036] S36. Define time-varying gain: Among them, q1(t) min is the lower bound of q1(t);

[0037] S37. Design the Lyapunov function as follows:

[0038]

[0039] And the derivative of the Lyapunov function designed in step S37 is obtained:

[0040]

[0041] S38. According to step S36 and step S37, the system control rate and the adaptive rate are designed as follows:

[0042]

[0043] in,

[0044] Further, step S4 specifically includes:

[0045] S41, the Lyapunov function designed in step S32 is used to obtain the quantization part, as follows:

[0046]

[0047] Pick Among them, l>0;

[0048] S42, will Substituting in, according to Young's inequality, we get:

[0049]

[0050] S43. Due to get make Thus we get according to And got

[0051] S44. From step S35 and step S37, we obtain:

[0052]

[0053] S45, according to Young's inequality and inequality get:

[0054]

[0055]

[0056] Among them, λ max (·) represents the maximum eigenvalue of ·;

[0057] S46, take Then we have:

[0058]

[0059] The control parameters are selected as follows:

[0060]

[0061] Among them, l represents the parameter to be designed;

[0062] S47. Due to Thus we get Substituting in:

[0063]

[0064] S48, by It can be seen that:

[0065] There is a maximum value, recorded as Q; select Therefore, we have:

[0066]

[0067] S49. Comprehensive consideration, due to Then V is bounded, and solving the above inequality yields:

[0068]

[0069] Obviously, all signals in the closed-loop system are semi-globally bounded when hour Then t→∞,

[0070] Compared with the prior art, the present invention has the following advantages:

[0071] 1. The present invention provides an unmanned surface vehicle adaptive neural network heading tracking control method with actuator input quantization and ship steering gear dynamic characteristics, and takes into account the characteristics of the ship rudder servo system with input quantization, solving the problem that the general ship steering gear system cannot achieve step steering.

[0072] 2. The adaptive neural network heading tracking control method for unmanned surface vehicles provided by the present invention introduces a linear input quantization analysis model, which ensures that the controller design does not rely on any a priori known quantization parameter information. Compared with the fixed quantization parameters used in the controller design process of the traditional quantization control method, the control method of the present invention has greater versatility and adaptability.

[0073] 3. The present invention provides an adaptive neural network heading tracking control method for unmanned surface vehicles, which adopts a dynamic surface control (DSC) algorithm to solve the "differential explosion" problem existing in traditional inversion control. In addition, it can achieve effective smoothing of quantized signals.

[0074] Based on the above reasons, the present invention can be widely promoted in fields such as artificial intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0076] Figure 1 The figure is a flow chart of the method of the present invention.

[0077] Figure 2 The simplified block diagram of the adaptive neural network control system with actuator input quantization and ship steering gear dynamic characteristics proposed by the present invention.

[0078] Figure 3 The tracking result diagram of the USV heading angle, USV bow angular velocity and USV rudder angle provided in the embodiment of the present invention.

[0079] Figure 4 The USV heading angle, USV bow angular velocity and USV rudder angle tracking error result diagram provided by the embodiment of the present invention.

[0080] Figure 5 A comparison diagram of control input curves provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0081] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0082] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0083] like Figure 1 As shown, the present invention provides an unmanned surface vehicle adaptive neural network heading tracking control method with actuator input quantization and ship steering gear dynamic characteristics, comprising:

[0084] S1. According to the surrounding environment and the sea conditions of other ships around, a third-order response unmanned surface vehicle heading control mathematical model is established;

[0085] S2, introduce the linear input quantification analysis model to quantify the control input in the control system;

[0086] S3. Based on the dynamic surface control algorithm, design the Lyapunov function to obtain the heading controller of the unmanned surface vehicle;

[0087] S4. Based on Lyapunov stability theory, the stability of the designed unmanned surface vehicle adaptive neural network heading tracking control system with input quantization is proved when there is no need for fixed prior information of quantization parameters, and all signals in the closed-loop system are semi-globally uniformly bounded. The tracking error can be made arbitrarily small by adjusting the controller parameters.

[0088] In specific implementation, as a preferred embodiment of the present invention, step S1 specifically includes:

[0089] S11. Construct a ship nonlinear Nomoto model and a ship steering gear dynamic characteristics model as follows:

[0090]

[0091] Among them, ψ represents the ship's heading angle, δ is the rudder angle, K and T represent the ship's motion maneuverability index, α and β represent nonlinear parameters, and δ E Indicates the command rudder angle issued by the servo, K E Indicates the servo control gain, T E Indicates the servo time constant;

[0092] S12. Introduce input quantization and construct a third-order response mathematical model of the unmanned surface vehicle as follows:

[0093]

[0094] in, ω1 and ω2 both represent unknown disturbances caused by wind, current and waves.

[0095] In specific implementation, as a preferred embodiment of the present invention, step S2 specifically includes:

[0096] S21. For the control input after considering the input quantization in the ship model, let Q(u) = q1(t)u + q2(t), and take:

[0097]

[0098] S22. Since the sign of the quantization process remains unchanged, q1(t) is unknown. From the formula in step S21, we know that q1(t)>0; when |u(t)|<a, Q(u) is bounded, q1(t)=1, then q2(t) is bounded, so we take

[0099] In specific implementation, as a preferred embodiment of the present invention, step S3 specifically includes:

[0100] S31, define heading error as: S1 = x1-x ld , S2=x2-x 2d , S3=x3-x 3d , and respectively differentiate the heading error to obtain:

[0101]

[0102] S32. Define the Lyapunov function as follows:

[0103]

[0104] Among them, y i represents the error of the virtual control item,

[0105] S33. Define a first-order low-pass filter as follows:

[0106]

[0107] in,

[0108] S34. Since f(x,t) is unknown, a RBF neural network is constructed to approximate the unknown function of the system model. There exists an ideal weight vector θ * ∈R N , so that the neural network θ *T h(x) is close enough to the given function f and the absolute value of the approximation error is no greater than σ M ,Right now:

[0109] f(x)=θ *T h(x)+ε *

[0110] S35. According to Young's inequality, the derivatives of V1 and V2 in step S32 are obtained:

[0111]

[0112] S36. Define time-varying gain: Among them, q1(t) min is the lower bound of q1(t);

[0113] S37. Design the Lyapunov function as follows:

[0114]

[0115] And the derivative of the Lyapunov function designed in step S37 is obtained:

[0116]

[0117] S38. According to step S36 and step S37, the system control rate and the adaptive rate are designed as follows:

[0118]

[0119] in,

[0120] In specific implementation, as a preferred embodiment of the present invention, step S4 specifically includes:

[0121] S41, the Lyapunov function designed in step S32 is used to obtain the quantization part, as follows:

[0122]

[0123] Pick Among them, l>0;

[0124] S42, will Substituting in, according to Young's inequality, we get:

[0125]

[0126] S43. Due to get make Thus we get according to And got

[0127] S44. From step S35 and step S37, we obtain:

[0128]

[0129] S45, according to Young's inequality and inequality get:

[0130]

[0131] Among them, λ max (·) represents the maximum eigenvalue of ·;

[0132] S46, take 0<η0≤1.0, Then we have:

[0133]

[0134] The control parameters are selected as follows:

[0135]

[0136] Among them, l represents the parameter to be designed;

[0137] S47. Due to Thus we get Substituting in:

[0138]

[0139] S48, by It can be seen that:

[0140] There is a maximum value, recorded as Q; select Therefore, we have:

[0141]

[0142] S49. Comprehensive consideration, due to Then V is bounded, and solving the above inequality yields:

[0143]

[0144] Obviously, all signals in the closed-loop system are semi-globally bounded when hour Then t→∞,

[0145] Example

[0146] In order to verify the effectiveness of the solution of the present invention, this embodiment uses MATLAB to perform computer simulation research, and the parameters are set as follows:

[0147] The simulation object is the "Lanxin" unmanned surface ship of Dalian Maritime University. The parameters of the unmanned ship are as follows: length 7.02m, width 2.6m, speed 35kn, full load draft 0.32m, full load displacement 2.73m 3 , square coefficient 0.6976. Adopt Norrbin motion model, take model parameters K = 0.71, T = 0.32, take nonlinear coefficient α = 1, β = 0.001.

[0148] Set the ideal heading of the unmanned ship to φ d =sin(0.1*t)+cos(0.2*t), the controller designed in step S3 is used to control the unmanned surface vehicle, wherein the design parameters are as follows: The quantization parameter is K=8.

[0149] The experimental results are as follows Figure 3-5 As shown, Figure 3 The tracking results of the heading angle, bow angular velocity and rudder angle of the unmanned surface vehicle are displayed. It can be seen that the controller designed in the present invention has good control performance and can realize ship heading tracking relatively quickly. Figure 4 are the ship heading angle tracking error, the ship bow angular velocity tracking error and the ship rudder angle tracking error. It can be seen that the tracking error can quickly converge to a smaller residual set, which further verifies the effectiveness of the controller designed in this invention. Figure 5 In order to consider the control input curves with and without actuator input quantization, Figure 5 It can be seen that the input quantization process has the effect of reducing the signal transmission burden, lowering the execution frequency of the actuator, and making the process closer to the control law of the multi-stage servo system in navigation practice.

[0150] The simulation results verify the effectiveness of the adaptive neural network controller used in the heading tracking task. The control system can effectively cope with changes and interference in the external environment and ensure the stability of heading control. By introducing input quantization technology, not only the communication burden of the unmanned surface vehicle is reduced, but also the execution frequency of the actuator is reduced. From the practical application point of view, in order to obtain good heading control performance, the present invention also takes into account the characteristics of the ship's steering gear servo system. Finally, the effectiveness of the unmanned surface vehicle adaptive neural network heading tracking control method proposed by the present invention with actuator input quantization and ship steering gear dynamic characteristics is further verified.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An adaptive neural network heading tracking control method for an unmanned surface vehicle, characterized in that: include: S1. According to the surrounding environment and the sea conditions of other ships around, a third-order response unmanned surface vehicle heading control mathematical model is established; S2, introduce the linear input quantification analysis model to quantify the control input in the control system; S3. Based on the dynamic surface control algorithm, design the Lyapunov function to obtain the heading controller of the unmanned surface vehicle; S4. Based on Lyapunov stability theory, the stability of the designed unmanned surface vehicle adaptive neural network heading tracking control system with input quantization is proved when there is no need for fixed prior information of quantization parameters, and all signals in the closed-loop system are semi-globally uniformly bounded. The tracking error can be made arbitrarily small by adjusting the controller parameters.

2. The method for controlling the heading tracking of an unmanned surface vehicle using an adaptive neural network according to claim 1, characterized in that: Step S1 specifically includes: S11. Construct a ship nonlinear Nomoto model and a ship steering gear dynamic characteristics model as follows: Among them, ψ represents the ship's heading angle, δ is the rudder angle, K and T represent the ship's motion maneuverability index, α and β represent nonlinear parameters, and δ E Indicates the command rudder angle issued by the servo, K E Indicates the servo control gain, T E Indicates the servo time constant; S12. Introduce input quantization and construct a third-order response mathematical model of the unmanned surface vehicle as follows: in, ω1 and ω2 both represent unknown disturbances caused by wind, current and waves.

3. The method for controlling the heading tracking of an unmanned surface vehicle using an adaptive neural network according to claim 1, characterized in that: Step S2 specifically includes: S21. For the control input after considering the input quantization in the ship model, let Q(u) = q1(t)u + q2(t), and take: S22. Since the sign of the quantization process remains unchanged, q1(t) is unknown. From the formula in step S21, we know that q1(t)>0; when |u(t)|<a, Q(u) is bounded, q1(t)=1, then q2(t) is bounded, so we take 4. The method for controlling the heading tracking of an unmanned surface vehicle using an adaptive neural network according to claim 1, characterized in that: Step S3 specifically includes: S31, define heading error as: S1 = x1-x ld , S2=x2-x 2d , S3=x3-x 3d , and respectively differentiate the heading error to obtain: S32. Define the Lyapunov function as follows: Among them, y i represents the error of the virtual control item, i=2,3; S33. Define a first-order low-pass filter as follows: in, S34. Since f(x,t) is unknown, an RBF neural network is constructed to approximate the unknown function of the system model. There exists an ideal weight vector θ * ∈R N , so that the neural network θ *T h(x) is close enough to the given function f and the absolute value of the approximation error is no greater than σ M ,Right now: f(x)=θ *T h(x)+e * S35. According to Young's inequality, the derivatives of V1 and V2 in step S32 are obtained: S36. Define time-varying gain: Among them, q1(t) min is the lower bound of q1(t); S37. Design the Lyapunov function as follows: And the derivative of the Lyapunov function designed in step S37 is obtained: S38. According to step S36 and step S37, the system control rate and the adaptive rate are designed as follows: in, 5. The method for controlling the heading tracking of an unmanned surface vehicle using an adaptive neural network according to claim 1, characterized in that: Step S4 specifically includes: S41, the Lyapunov function designed in step S32 is used to obtain the quantization part, as follows: Pick Among them, l>0; S42, will Substituting in, according to Young's inequality, we get: S43. Due to get make Thus we get according to And got S44. From step S35 and step S37, we obtain: S45, according to Young's inequality and inequality get: Among them, λ max (·) represents the maximum eigenvalue of ·; S46, take 0<η0≤1.0, Then we have: The control parameters are selected as follows: Among them, l represents the parameter to be designed; S47. Due to Thus we get Substituting in: S48, by i=1,2, we know: There is a maximum value, recorded as Q; select Therefore, we have: S49. Comprehensive consideration, due to Then V is bounded, and solving the above inequality yields: Obviously, all signals in the closed-loop system are semi-globally bounded when hour Then t→∞,

Images