Dispersive adaptive dynamic surface quantitative control method with unmodeled dynamic nonlinear system and application

Through the decentralized adaptive dynamic surface quantization control method, the problem of quadrotor drone ignoring dynamic uncertainty during the modeling process is solved, and higher control accuracy and tracking performance are achieved, verifying the feasibility of the method.

CN118938981APending Publication Date: 2024-11-12DATANG DONGBEI ELECTRIC POWER TESTING & RES INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410978570.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The quadrotor UAV ignores dynamic uncertain terms or uncertain parameters during system modeling, resulting in the control accuracy not reaching the expected effect.

Method used

The dispersed adaptive dynamic surface quantization control method with unmodeled dynamic nonlinear systems is adopted. By defining dynamic error surfaces, selecting Lyapunov functions, approximating unknown nonlinear functions using RBFNNs, and introducing hysteresis quantizers, designing The final law of control is to overcome dynamic uncertainty terms.

Benefits of technology

The control accuracy of the four-rotor drone is improved, ensuring that the system tracking error is within the preset tracking performance range, and the feasibility of the method is verified through simulation verification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118938981A_ABST
    Figure CN118938981A_ABST
Patent Text Reader

Abstract

The invention provides a decentralized self-adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system and application, and relates to the field of decentralized self-adaptive dynamic surface quantization control with an unmodeled dynamic nonlinear system. The problem that in the prior art, in the system modeling process of the quad-rotor unmanned aerial vehicle, some dynamic uncertain items or uncertain parameters are always ignored, so that the control precision of the quad-rotor unmanned aerial vehicle cannot achieve the expected effect is solved. The invention provides the following scheme: a decentralized adaptive neural network dynamic surface quantitative control method is adopted to carry out controller design on a nonlinear system with unmodeled dynamics and state delay, the assumption of predecessors on dynamic uncertain items in work is relaxed, and an error performance function is introduced, so that the reliability of the system is improved, and the reliability of the system is improved. The tracking error of the system reaches the preset tracking performance, the provided method is verified by using the Qball-X4 unmanned aerial vehicle system model, and the method is also suitable for the simulation verification of the Qball-X4 unmanned aerial vehicle system on the controller design.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of decentralized adaptive dynamic surface quantization control with unmodeled dynamic nonlinear systems, and in particular to a decentralized adaptive dynamic surface quantization control method and application with unmodeled dynamic nonlinear systems. Background Art

[0002] There are many uncertainties in control systems, such as unmodeled dynamics, external or internal disturbances, etc. In the prior art, an observer is designed to estimate the system state to solve the unmodeled dynamics and state uncertainty. In an in-depth study of the prior art, an available dynamic signal is introduced to deal with the unmodeled dynamics, and an adaptive backstepping method is used to design the controller. In addition, it can also be solved by describing the Lyapunov function. A robust adaptive control method based on a neural network is proposed for a class of state-delay nonlinear strict feedback systems. Therefore, some tracking control methods are proposed for nonlinear time-delay systems with unmodeled dynamics, and good results have been achieved in the controller. In addition, some scholars have constructed controllers for a class of random nonlinear systems with state and input unmodeled dynamics through adaptive neural control schemes.

[0003] However, in the system modeling process of quadrotor drones, some dynamic uncertainties or uncertain parameters are always ignored, which makes the control accuracy of the quadrotor drone fail to achieve the expected effect. Summary of the invention

[0004] The present invention aims to solve the problem that in the prior art, some dynamic uncertainties or uncertain parameters are always ignored during the system modeling process of a quad-rotor UAV, thereby causing the control accuracy of the quad-rotor UAV to fail to achieve the expected effect.

[0005] To solve the above technical problems, the present invention is implemented by the following technical solution: a decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system, the control method comprising the following steps:

[0006] The step of determining a dynamic error surface is to define at least one dynamic error surface:

[0007] s i,1 =x i,1 -y r1 (40)

[0008] Derivative of at least one dynamic error surface yields:

[0009]

[0010] Controller setup steps to select the final candidate Lyapunov function:

[0011]

[0012] In the formula yes The estimated value of The derivative is:

[0013]

[0014]

[0015] Substituting formula (44) and formula (45) into formula (43), we obtain:

[0016]

[0017] Use RBFNNs to approximate the unknown nonlinear function and get

[0018]

[0019] in,

[0020] And satisfy the following formula:

[0021]

[0022] Add a hysteresis quantizer step to convert Q(u i ) is expressed as

[0023] Q(u i )=q1(t)u i +q2(t) (49)

[0024] Substituting formula (48) and formula (49) into formula (46), we obtain:

[0025]

[0026] Among them, Through formula (51), the final control law is obtained

[0027]

[0028] Parameter design steps, parameter adjustment rate is expressed as:

[0029]

[0030] In the formula, is a positive design parameter, for The estimated parameters of The estimated value of is:

[0031]

[0032] in, Design parameters for the system.

[0033] Furthermore, a preferred embodiment is provided, in which the preset conditions for using RBFNNs to approximate an unknown nonlinear function are:

[0034]

[0035] αγ are design parameters, is the jth error surface.

[0036] Further, a preferred embodiment is provided, wherein the controller setting step is implemented based on the following formula:

[0037]

[0038] Furthermore, a preferred implementation is provided, wherein each step of the decentralized adaptive dynamic surface quantization control method is implemented using RBFNN and a low-pass filter.

[0039] Furthermore, a preferred embodiment is provided, wherein the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system also includes a stability and transient performance analysis step.

[0040] Solution 2: The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system described in Implementation Method 1 is used as an embedded program in the simulation verification system for the controller design of the Oball-x4 UAV system.

[0041] Solution three: a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the method described in any one of Solution one.

[0042] Solution 4: A computer device includes a memory and a processor, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes any one of the methods described in Solution 1.

[0043] The present invention is beneficial in that:

[0044] The present invention adopts a decentralized adaptive neural network dynamic surface quantization control method to design a controller for a nonlinear system with unmodeled dynamics and state delays, relaxes the assumptions about dynamic uncertainties made by predecessors in their work, and introduces an error performance function to enable the tracking error of the system to reach a preset tracking performance. In the simulation experiment, the Qball-X4 UAV system model is used to verify the proposed method, and the simulation results verify the feasibility of the method.

[0045] The present invention is also applicable to the simulation verification of controller design in the Qball-X4 UAV system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a schematic diagram of the simulation of the Qball-X4 UAV position subsystem described in Implementation Example 9.

[0047] Figure 2 Schematic diagram of the change of X position tracking performance and tracking error described in Implementation Example 9.

[0048] Figure 3 Schematic diagram of the change of Y position tracking performance and tracking error described in Implementation Example 9.

[0049] Figure 4 Schematic diagram of the change of Z position tracking performance and tracking error described in Implementation Example 9.

[0050] Figure 5 Schematic diagram of the changes in pitch angle tracking performance and tracking error described in Implementation Example 9.

[0051] Figure 6 Schematic diagram of the changes in the roll angle tracking performance and tracking error described in Implementation Example 9.

[0052] Figure 7 Schematic diagram of the change of yaw angle tracking performance and tracking error described in Implementation Example 9.

[0053] Figure 8 This is a schematic diagram for comparing tracking errors of the position subsystem described in Implementation Example 9.

[0054] Fig. 9 This is a schematic diagram of the comparison of tracking errors of the posture subsystem described in Implementation Mode 9.

[0055] Fig.10 Schematic diagram of the spiral increase in the expected tracking performance of the quadrotor drone described in Implementation Example 9. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the implementation methods of the present application clearer, the technical solutions in the implementation methods of the present application will be clearly and completely described below in conjunction with the drawings in the implementation methods of the present application. Obviously, the described implementation methods are only part of the implementation methods of the present application, not all of the implementation methods.

[0057] Embodiment 1: This embodiment provides a decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system, and the control method includes the following steps:

[0058] The step of determining a dynamic error surface is to define at least one dynamic error surface:

[0059] s i,1 =x i,1 -y r1 (40)

[0061] Derivative of at least one dynamic error surface yields:

[0062]

[0063] Controller setup steps to select the final candidate Lyapunov function:

[0064]

[0065] In the formula yes The estimated value of The derivative is:

[0066]

[0067]

[0068] Substituting formula (44) and formula (45) into formula (43), we obtain:

[0069]

[0070] Use RBFNNs to approximate the unknown nonlinear function and get

[0071]

[0072] in, And satisfy the following formula:

[0073]

[0074] Add a hysteresis quantizer step to convert Q(u i ) is expressed as

[0075] Q(ui )=q1(t)u i +q2(t) (49)

[0076] Substituting formula (48) and formula (49) into formula (46), we obtain:

[0077]

[0078] Among them, Through formula (51), the final control law is obtained

[0079]

[0080] Parameter design steps, parameter adjustment rate is expressed as:

[0081]

[0082] In the formula, is a positive design parameter, for The estimated parameters of The estimated value of is:

[0083]

[0084] in, Design parameters for the system.

[0085] Embodiment 2: This embodiment further limits the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system described in Embodiment 1. The preset conditions for using RBFNNs to approximate the unknown nonlinear function are:

[0086]

[0087] αγ are design parameters, is the jth error surface.

[0088] Embodiment 3: This embodiment further limits the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system described in Embodiment 1. The controller setting step is implemented based on the following formula:

[0089]

[0090] Implementation method 4: This implementation method further limits the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system described in implementation method 1. The decentralized adaptive dynamic surface quantization control method is implemented using RBFNN and a low-pass filter in each step.

[0091] Implementation method five: This implementation method is a further limitation of the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system described in implementation method one. The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system also includes stability and transient performance analysis steps.

[0092] Implementation method six: This implementation method proposes a decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system as an embedded program of the Oball-x4 UAV system for simulation verification of controller design.

[0093] Implementation method seven: This implementation method proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system described in any one of the implementation methods one.

[0094] Embodiment 8. This embodiment proposes a computer device, including a memory and a processor, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system as described in any one of Embodiments 1.

[0095] Implementation method 9: See Figures 1 to 10 This embodiment is described. Embodiment 9 is used to explain Embodiments 1 to 8.

[0096] In addition, there is coupling between the position system and attitude system of the quadrotor, which belongs to a decentralized control system. Therefore, the adaptive decentralized control method is used to design the controller. In order to increase the network communication rate, a quantizer needs to be introduced. The quantizer can convert the continuous signal of the system into a discrete signal, thereby improving the control accuracy of the system.

[0097] For a class of pure feedback nonlinear systems with unmodeled dynamics, a decentralized neural network dynamic surface quantization control method is used to design the controller. The main work is as follows:

[0098] 1) For unknown nonlinear functions and dynamic uncertainties, RBFNNs are used to approximate them, and their estimates are expressed in the form of norms.

[0099] 2) By introducing a hysteresis quantizer, the controller is designed using a decentralized adaptive dynamic surface quantization control method to reduce system communication time, improve control accuracy, overcome parameter uncertainty and other issues, and introduce an error conversion function to keep the system tracking error within a pre-given tracking performance range. The feasibility of the method is verified through a quadrotor UAV system simulation.

[0100] First, consider a nonlinear pure feedback control system with dynamic uncertainty:

[0101]

[0102] in, is the state variable of the system; is the system's time delay state quantity; similarly, we can get is the delay state; Δ i,j (x,z,t) is the system dynamic uncertainty or dynamic disturbance term; z is the unmodeled dynamic; is the interconnection term between the i-th subsystem and the j-th subsystem; is the control system output; Q(u i ) is the quantized input signal; where Δ i,j (x,z,t) and q i (z i ,x i ) satisfies Lipschitz uniform continuity.

[0103] Apply the mean value theorem to formula 1

[0104]

[0105] When 0 <v i <1, yes At the same time 0 <v i,ni When <1, we can get Rewrite Formula 2 into the following form

[0106]

[0107] In order to simplify the analysis process, we define

[0108]

[0109] where g i,j (·), is an unknown continuous function. Substituting formula (3) into formula (1), we get

[0110]

[0111] Assumption 1: Reference signal y r It is known that when t≥0, within a compact set.

[0112] Assumption 2: Define g i (·), i=1,…,n are known. To facilitate the design of the controller, let g i (·)>0 and there exists g min ,gmax Satisfy min ≤|g(·)|≤g max .

[0113] Assumption 3: For the dynamic uncertainty term Δ i,j (x,z,t),i=1,...n,j=1,...,n i have

[0114]

[0115] Among them, φ i,1 (·),φ i,2 (·) are all unknown smooth functions. Assume that the initial φ i,2 (·)=0.

[0116] Assumption 4: The unmodeled dynamic terms satisfy the exp-ISpS stability condition, so the system is defined as The Lyapunov function V(z) that satisfies the exp-ISpS condition i )

[0117]

[0118] where α i,1 (·),α i,2 (·) is a type K ∞ function, λ(·) is a known smooth function.

[0119] Assumption 5: For the coupled time-delay terms in the interconnected terms and the unknown functions, we have

[0120]

[0121] If System V is System The exp-ISpS lyapunov function, that is, formula (1) and (10) exist for any constant in (0,c) Under any initial condition t0>0, any z0=z(t0) and V0>0, the function exist Function D(t0,t), and signal v>0, define t≥t0 and there exists a function Make it

[0122] t≥t i.0 +T i,0 ,D i,j (t0,t)=0,V i,j (z i )≤v i,j (t)+D(t0,t) for all t≥t0,

[0123] Assumptions Therefore, the existential Establishment

[0124] Secondly, the error transfer function

[0125] make

[0126] e i,j =x i,j -y r.i (10)

[0127] where y r.i is a given reference signal, defining the performance function is a strictly increasing function, for all t>0

[0128]

[0129] Where 0<σ<1, In the formula is the maximum allowable value of the steady-state tracking error, which is an equivalent function of formula (11), so there is an error performance function

[0130]

[0131] Where s1 is the conversion error, Φ(s1) is a strictly non-decreasing function, and its inverse function satisfies the following formula:

[0132]

[0133] From formula (14), it can be concluded that if s i,1 ∈L ∞ , then formula (13) holds, and let According to formula (12), we can get Formula (11) holds true, so in order to achieve the pre-specified tracking performance, only s i,1 ∈L ∞ , and since Φ(s1) is a monotonically increasing function, the error conversion form of the first dynamic surface is obtained

[0134]

[0135] The following are the controller design steps:

[0136] Step 1: Define the first dynamic error surface

[0137] s i,1 =x i,1 -y r1 (16)

[0138] Among them, y r1 For a given reference trajectory

[0139] Taking the derivative of its dynamic error surface, we can get

[0140]

[0141] Where b i,1 >0,e i,1 > 0. The coupled time-delay terms of the unknown function are all treated using the theory in Section 2.3. Next, select the Lyapunov function

[0142]

[0143] Where λ0>0, in the subsequent controller design process, formula (19) is applied to each step of the dynamic surface design

[0144]

[0145] Where α i,1 >0, and from assumptions 1-4 we can get

[0146]

[0147] For the error transfer function and the unmodeled dynamic terms, the following equation is satisfied

[0148]

[0149] where α i,1 >0,α i,2 >0, where φ i,1 ,φ i,2 It has been defined in Assumption 3, and can be obtained by formula (21)

[0150]

[0151] Where α γ Always positive, v i,1 is a positive parameter, δ is a positive constant and can be given arbitrarily by the designer, x i,2d is the virtual control law of the first dynamic error surface, d i,1max is a smooth function and monotonically increasing. In the next step, d i,jmax i=1,…,nj=1,…,n can also be obtained in a similar way, that is,

[0152] In an embodiment, RBFNNs are used to approximate unknown nonlinear functions and coupled time-delay functions.

[0153] α γ is the design parameter, is the jth error surface, which will be reflected in the subsequent design process)

[0154]

[0155] in ξ i,1 It is an arbitrary design parameter order

[0156]

[0157] Where g m =min{g min}, the formula is established

[0158]

[0159] where v i,1 is the design parameter. Substituting equation (23) into equation (22) and considering equation (25), we can get

[0160]

[0161] make

[0162] In order to make equation (27) stable, the virtual control of the first step is selected as

[0163]

[0164] where k i,1 is a positive design parameter, is the neural network θ * The estimated value of will be given uniformly in the last step of controller design.

[0165] Let x i,2d Through a first-order low-pass filter, we get the new variable z i,2

[0166] τ i,j z i,j +z i,j =x i,j+1d; z i,j (0) = x i,j+1d; (28)

[0167] where τ i,j is the time constant of the first-order low-pass filter.

[0168] Step j (2≤j≤n-1): Define the jth error surface

[0169] s i,j =x i,j -z i,j (29)

[0170] Derivative of the jth error surface

[0171]

[0172] Select the following Lyapunov function

[0173]

[0174] V i,j The derivation can be obtained

[0175]

[0176] According to formula (21), we can get

[0177]

[0178] in It has been defined in Assumption 3. According to formula (21), substituting formula (33) into formula (32) yields

[0179]

[0180] Similar to formula (23), we have

[0181]

[0182] In the formula ξ i,j is any positive design parameter, with the formula

[0183]

[0184] Among them, v i,j , ξ is the design parameter, considering equation (4-35), substituting equation (4-36) into (4-34) we have

[0185]

[0186] In the formula In order to make equation (37) stable, the virtual control of the jth step is given

[0187]

[0188] where k i,j is a positive design parameter, let x i,j+1d Through a first-order low-pass filter, we get the new variable z i,j+1

[0189] τ i,j+1 z i,j+1 +z i,j+1 =x i,j+1d; z i,j+1 (0) = x i,j+1d; (0) (39)

[0190] where τ i,j+1 is the jth (j=1,…,n-1) first-order low-pass filter time constant.

[0191] Step n: The control method comprises the following steps:

[0192] The step of determining a dynamic error surface is to define at least one dynamic error surface:

[0193]

[0194] Derivative of at least one dynamic error surface yields:

[0195]

[0196] Controller setup steps to select the final candidate Lyapunov function:

[0197]

[0198] In the formula yes The estimated value of The derivative is:

[0199]

[0200] Substituting formula (44) and formula (45) into formula (43), we obtain:

[0201]

[0202] Use RBFNNs to approximate the unknown nonlinear function and get

[0203]

[0204] in,

[0205] And satisfy the following formula:

[0206]

[0207] Add a hysteresis quantizer step to convert Q(u i ) is expressed as

[0208] Q(u i )=q1(t)u i +q2(t) (49)

[0209] Substituting formula (48) and formula (49) into formula (46), we obtain:

[0210]

[0211] Among them, Through formula (51), the final control law is obtained

[0212]

[0213] Parameter design steps, parameter adjustment rate is expressed as:

[0214]

[0215] In the formula, is a positive design parameter, for The estimated parameters of The estimated value of is:

[0216]

[0217] in, Design parameters for the system.

[0218] In this section, we will introduce the stability and transient performance analysis of the decentralized adaptive dynamic surface quantization control method, define

[0219]

[0220] where x i,2d , x i,j+1d Given by (27), (38), by (28), (39), (54), (55), combined with formula (39)

[0221]

[0222] Conclusion

[0223]

[0224] in is a continuous function.

[0225] Theorem 1: For the closed-loop system composed of assumptions 1-5, the controller design satisfies formula (57) and formula (58), then the Lyapunov function

[0226]

[0227] Where V i,j Defined by formula (31) and formula (42) respectively, for a given constant ε i,1 , we can make formula (23), formula (35), and formula (47) valid, and let them all be and In the compact set of, for a given design parameter p, if V(0)<p, then there exists k i,j ,τ i,j ,σ q ,σ v All signals in the closed-loop system converge to a certain neighborhood, and by reasonably selecting design parameters, they can converge to any small value within the residual set.

[0228] Proof: Taking the derivative of V, we get

[0229]

[0230] From (29), (40), we have

[0231] x i,j =s i,j +y i,j +x i,j+1d ,i=1,...,n,j=1,...,n (61)

[0232] Substituting (22), (38) and (61) into (26) and (37), we have

[0233]

[0234] Among them for There is a formula

[0235]

[0236] Substituting (52) into (50), we obtain

[0237]

[0238] Considering equation (58), substituting (62)-(64) into (60), the following equation holds:

[0239]

[0240] By Assumption 1, we define a compact set

[0241]

[0242] It belongs to Among them, Π1×Π2 also belongs to Therefore |B i,j |There is a maximum value M in Π1×Π2 i,j , according to the formula and parameter adjustment rate

[0243]

[0244] According to formula (63), we have

[0245]

[0246] For the parameter μ, we have the following formula

[0247]

[0248] where i = 1, ..., n, j = 1, ..., n, let

[0249]

[0250] in Substitute (70)(71) into (69)

[0251]

[0252] Where -α i0 is a negative constant, C is a constant term, and the definition is

[0253]

[0254] Due to s i,1 Uncertain, so consider two situations

[0255] Case 1: Assume For any parameter α in (73) γ ,have

[0256] x i,1 =s i,1 -y r,1 (74)

[0257] s i,1 is bounded, and from equation (72) we can get

[0258]

[0259] set up Then there is V=p, that is, V≤p is an invariant set, that is, V(0)≤p, then for all t≥0, solve formula (75)

[0260]

[0261] Similar to (76), we have

[0262]

[0263] Case 2: Assume Formula [(1-2tanh 2 (s i,1 / α γ ))]>0, from (75) we have

[0264]

[0265] have

[0266] α i0 >C / p (79)

[0267] Then there is V=p, that is, V<p is an invariant set, that is, V(0)≤p, then for all t≥0, solve formula (75)

[0268]

[0269] This means that t→∞ V(t)=(C / α i0 ).

[0270] According to these two cases, there is s i,j , y i,j are uniformly bounded. In addition, appropriate selection of design parameters Make α i0 can be arbitrarily large, and according to formula (76), the tracking error obtained by formula (80) can be arbitrarily small.

[0271] The neural network dynamic surface method proposed in this embodiment simplifies the steps of designing the controller; however, since RBFNN and low-pass filter are used in each design step, the system can only achieve semi-global consistent stability. In addition, as long as α i0 If is large enough, the time constant of the low-pass filter will be arbitrarily small, however, this may result in a higher amplitude control signal. Therefore, from an engineering perspective, a trade-off must be made between tracking performance and the amplitude of the control signal in order to obtain better tracking performance indicators.

[0272] In order to ensure the tracking performance of the system, the Qball-X4 UAV system is used to simulate and verify the controller design. The basic parameters of Qball-X4 have been given in the simulation analysis in Chapter 3 and will not be repeated in this section. The simulation content is divided into two parts: the position subsystem and the attitude subsystem. The following is a detailed description of the parameter selection of the controller design and the simulation results obtained through the UAV system.

[0273] Consider the position subsystem of a quadrotor drone with unmodeled dynamics and time delays:

[0274]

[0275] in are the three control inputs of the position subsystem, [Δ X ,Δ Y,Δ Z ] is a dynamic uncertainty term, f X (x) = f(x X,1 -x X,1 (t-2)) / m,f Y (x) = f(x Y,1 -x Y,1 (t-2)) / m,f Z (x) = f(x Z,1 -x Z,1 (t-2)) / m is the unknown time lag function, Q(U X ),Q(U Y ),Q(U Z ), is the quantitative input of the position subsystem, m = 20 is the design value, so the following parameters are set:

[0276] First, define the dynamic uncertainty term Δ X (x, t, z) = 0.01sin(t) + z X , Δ Y (x, t, z) = 0.01sin(t) + z Y , Δ Z (x, t, z) = 0.01sin(t) + z Z

[0277] Combine equation (1) and assumption 4 to select an observable dynamic signal:

[0278]

[0279] Select the error performance function:

[0280] F1(s1)=2 / parctan(s1), F2(s2)=2 / parctan(s2), F3(s3)=2 / parctan(s3)

[0281] And it meets the preset tracking performance v1(t)=v2(t)=v3(t)=0.8e -0.2t +0.015, and then we get s1 = tan(pe1(t) / 2v1(t)), s2 = tan(pe2(t) / 2v2(t)), s3 = tan(pe3(t) / 2v3(t))

[0282] Select reference signal [x0, y0, z0] = [cost, sint, 0.5t], select virtual control law and actual control law parameters, c1 = 7, c2 = 9, c3 = 6, c4 = 8, c5 = 12, c6 = 10, select filter time τ1 = τ2 = τ3 = 0.01, and select neural network Gaussian function: in, Therefore, 21 center points are selected, scattered in [-2, +2] × [-2, +2] × [-2, +2], the base width is selected as h1 = h2 = h3 = 1, j = 1, K, 21; the adaptive adjustment law is selected as σ i =0.001(i=1,2,3), based on the above definition and parameter selection, the simulation results are as follows Figures 2 to 4 shown.

[0283] Figures 5 to 7 It is shown in the figure that the designed controller tracks the reference signal well, overcomes the influence of parameter uncertainty and coupling time lag on the system, ensures that the tracking error between the system output y=x1 and the given signal is no more than 0.015, and within the preset tracking performance ω i (t) = 0.8e -0.2t +0.015i=1,2,3 neighborhood fluctuations, and due to the introduction of hysteresis quantizer, it is more effective to solve the delay problem and enhance the communication speed between each subsystem.

[0284] The feasibility of the controller design can be obtained through simulation experiments on the position subsystem. Figures 8 to 9 It means that all its control inputs meet the process design standards. Since there is a coupling relationship between the position subsystem and the attitude subsystem in the quadrotor UAV, the reference signal of the attitude subsystem is affected by the control signal of the position subsystem, that is, the reference signal of the attitude subsystem needs to be solved through the control input of the position subsystem. In the following content, the attitude group system of the UAV system with unmodeled dynamics will be simulated and verified.

[0285] Simulation of Qball-X4 UAV attitude subsystem

[0286] Consider the UAV attitude subsystem model with unmodeled dynamics:

[0287]

[0288] Where Q(U1), Q(U2), Q(U3) represent the quantized control inputs of the roll angle (φ), pitch angle (θ), and yaw angle (ψ), respectively, and the time lag term is:

[0289] f1(x)=f(x1-x1(t-2)) / m, f2(x)=f(x2-x2(t-2)) / m,

[0290] f3(x)=f(x3-x3(t-2)) / m, where m=20 is the design value, [Δ φ ,Δ θ ,Δ ψ ] is the disturbance term of the attitude subsystem, so given the design parameters:

[0291] First, define the dynamic uncertainty term:

[0292] Δ φ (x, t, z) = 0.01sin(t) + z φ , Δ θ (x, t, z) = 0.01sin(t) + z θ Δ ψ (x, t, z) = 0.01sin(t) + z ψ ,

[0293] Choose an observable dynamic signal based on the above:

[0294]

[0295] Select the error performance function:

[0296] F1(s1)=2 / parctan(s1), F2(s2)=2 / parctan(s2), F3(s3)=2 / parctan(s3)

[0297] And it meets the preset tracking performance v1(t)=v2(t)=v3(t)=0.8e -0.2t +0.015, and then we get

[0298]

[0299] According to the explanation in Section 4.5.1, the reference signal of the attitude subsystem is given by the input signal of the position subsystem, so the reference signals of the pitch angle and roll angle are shown in formula (84):

[0300]

[0301] Where a = cos(x z,2 ),b=sin(x z,2 ),c=(U Y +U X ) / U Z ,d=a+b,let the yaw angle reference signal y ψ =0.

[0302] Therefore, the virtual control and actual control parameters are selected as follows: c1 = 8, c2 = 9, c3 = 6, c4 = 15, c5 = 13, c6 = 10, the filter time τ1 = τ2 = τ3 = 0.01, and the neural network Gaussian function is selected as follows: ψ1(ξ1) = exp[-(ξ1-ζ1) T (ξ1-ζ1) / η1 2 ], in, Therefore, 21 center points are selected, scattered in [-2, +2] × [-2, +2] × [-2, +2], the base width is selected as h1 = h2 = h3 = 1, j = 1, K, 21; the adaptive adjustment law is selected as σ i = 0.001 (i = 1, 2, 3), according to the reference signal and parameter selection defined above, from Figures 5 and 6 The tracking performance shows that its pitch angle and roll angle tracking performance meet the expected performance indicators, and its error tracking curve is also within the pre-set tracking performance, achieving good control effect.

[0303] From the perspective of the quantitative control input of the pitch angle and roll angle, the quantization error is very small, and the initial overshoot is around ±1°, which achieves the control purpose of the system and enables the quadcopter to achieve the expected tracking performance. Finally, the six degrees of freedom are comprehensively investigated through the spiral rise curve, and the Fig.10 This proves that this method can achieve good control effect.

[0304] In this embodiment, first, the quantitative control method makes flight control more precise and accurate by accurately quantifying various flight parameters of the drone, such as attitude angle, speed, acceleration, etc. This high-precision control strategy can significantly improve the stability and adaptability of the quadcopter drone in complex environments. In environments with strong winds or large airflow disturbances, the drone can still maintain a stable flight attitude.

[0305] Secondly, the quantitative control method also promotes the application of intelligent control algorithms in quadrotor drones. By introducing advanced control algorithms, such as sliding membrane variable structure control and nonlinear control, drones can better cope with problems such as parameter uncertainty and external disturbances. These intelligent control algorithms combined with quantitative control methods enable drones to adjust control strategies in real time during flight to ensure flight safety and efficiency.

[0306] By accurately quantifying the position and attitude information of drones and combining it with advanced navigation and positioning technologies, drones can achieve higher-level autonomous flight functions, such as autonomous obstacle avoidance, autonomous path planning, etc. At the same time, in terms of formation flight, the quantitative control method can ensure that each drone maintains an accurate relative position and attitude, thereby achieving a more coordinated formation flight effect.

[0307] Finally, the application of quantitative control methods has also promoted the application expansion of quadrotor drones in multiple fields. Whether in the fields of aerial photography, logistics distribution, search and rescue, agricultural plant protection, scientific research, exploration, military reconnaissance, etc., quantitative control methods enable drones to better meet the needs of various complex tasks.

[0308] This implementation adopts a decentralized adaptive neural network dynamic surface quantization control method to design a controller for a nonlinear system with unmodeled dynamics and state delays, relaxes the assumptions about dynamic uncertainties made by previous researchers, and introduces an error performance function to enable the system's tracking error to reach a preset tracking performance. In the simulation experiment, the Qball-X4 UAV system model is used to verify the proposed method, and the simulation results verify the feasibility of the method.

[0309] Those skilled in the art will appreciate that the above are only preferred embodiments of the present invention, and the features described in the various embodiments and / or claims of the present disclosure may be combined or combined in a variety of ways, even if such combinations or combinations are not explicitly described in the present disclosure. The above are not intended to limit the present invention, and the above are only preferred embodiments of the present invention, and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art may still modify the technical solutions described in the aforementioned embodiments, or perform equivalent substitutions on some of the technical features therein, and any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

[0310] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A decentralized adaptive dynamic surface quantization control method for an unmodeled dynamic nonlinear system, characterized in that: The control method comprises the following steps: The step of determining a dynamic error surface is to define at least one dynamic error surface: s i,1 =x i,1 -y r1 (40) Derivative of at least one dynamic error surface yields: Controller setup steps to select the final candidate Lyapunov function: In the formula yes The estimated value of The derivative is: Substituting formula (44) and formula (45) into formula (43), we obtain: Use RBFNNs to approximate the unknown nonlinear function and get in, And satisfy the following formula: (48) Add a hysteresis quantizer step and convert Q(u i ) is expressed as Q(u i )=q1(t)u i +q2(t) (49) Substituting formula (48) and formula (49) into formula (46), we obtain: Among them, Through formula (51), the final control law is obtained Parameter design steps, parameter adjustment rate is expressed as: In the formula, is a positive design parameter, for The estimated parameters of The estimated value of is: in, Design parameters for the system.

2. The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system according to claim 1, characterized in that: The preconditions for using RBFNNs to approximate unknown nonlinear functions are: α γ is the design parameter, is the jth error surface.

3. The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system according to claim 1, characterized in that: The process of the controller setting step is based on the following formula:

4. The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system according to claim 1, characterized in that: The decentralized adaptive dynamic surface quantization control method is implemented using RBFNN and a low-pass filter in each step.

5. The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system according to claim 1, characterized in that: The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system also includes stability and transient performance analysis steps.

6. The decentralized adaptive dynamic surface quantization control method with an unmodeled dynamic nonlinear system as described in claim 1 is used as an embedded program in a simulation verification system for controller design of an Oball-x4 unmanned aerial vehicle system.

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

8. A computer comprising a processor and a storage medium, wherein when the processor reads a computer program stored in the storage medium, the computer executes the method of claim 1.