Unmanned helicopter inverse optimization cooperative fault-tolerant control method based on composite learning strategy

Abstract

The invention is suitable for the technical field of unmanned helicopter control, and provides an unmanned helicopter inverse optimization cooperative fault-tolerant control method based on a composite learning strategy, and the method comprises the steps: unifying a six-degree-of-freedom dynamic model of each unmanned helicopter under the influence of an actuator fault, dividing the six-degree-of-freedom dynamic model into two subsystems, namely a position ring and an attitude ring, designing a fault estimation observer based on a composite learning method through a series-parallel estimation model in combination with a neural network, performing real-time estimation on a lumped actuator fault and a subsystem state, and designing an inverse optimization cooperative fault-tolerant control strategy based on interaction information and an inverse optimization theory for the two subsystems; and tracking of the position and the yaw angle of the virtual leader by the unmanned helicopter follower is realized. Therefore, the cooperative fault-tolerant controller designed according to the composite learning and inverse optimization theory has good fault compensation performance and strong robustness, the cost function is minimized, and the dual requirements of stability and cost reduction of a real-time unmanned helicopter system are met.

Description

Technical Field

[0001] This application belongs to the technical field of unmanned helicopter control, and particularly relates to an inverse optimization cooperative fault-tolerant control method for unmanned helicopters based on a composite learning strategy. Background Technique

[0002] As a representative of advanced aviation equipment, an unmanned helicopter mainly provides upward lift through the main rotor, and changes its attitude angle by changing the pitch of the main rotor and the tail rotor, so as to achieve characteristics such as vertical takeoff and landing, hovering in the air, and low-altitude and low-speed flight. Compared with other unmanned aerial vehicles, unmanned helicopters also have the following advantages: less affected by the site, strong environmental adaptability; strong load capacity, flexible maneuverability and long hovering time; strong concealment, and can fly at ultra-low altitude and close to the ground in special environments; high level of intelligence, and in the actual use process, in addition to the predetermined task objectives, automatic operation can also be achieved. Therefore, unmanned helicopters are widely used in civilian fields such as cargo transportation, as well as military fields such as reconnaissance and surveillance. Compared with a single-machine platform, the real-time communication between unmanned helicopters in a multi-aircraft formation can improve the amount and accuracy of the information obtained, and cooperate to complete the desired tasks. Based on this advantage of the unmanned helicopter formation, the multi-aircraft formation cooperative control technology has become one of the research hotspots in the international aviation community.

[0003] However, an unmanned helicopter formation is usually a nonlinear system with underactuation, strong coupling or uncertainty. During the process of the formation performing tasks, various unexpected factors will inevitably cause components such as actuators and sensors to fail and deviate from the current task, affecting the performance and stability of a single unmanned helicopter within the formation. Moreover, the unmanned helicopters communicate with each other through a communication network, and a single-machine failure or network topology failure may be transmitted to neighboring unmanned helicopters, making it difficult to achieve the formation cooperative control goal. Therefore, if an unmanned helicopter formation system is to operate safely and reliably, a corresponding fault-tolerant control mechanism needs to be equipped to improve the fault tolerance of the formation and ensure that the formation can maintain overall stability even when suffering from faults.

[0004] As unmanned helicopters are becoming more and more intelligent and autonomous, regarding unmanned helicopters as intelligent agents and using advanced intelligent control technologies to solve the multi-aircraft formation cooperative fault-tolerant control problem has gradually been pushed to the forefront of the research field.

[0005] In addition, inverse optimal control can avoid solving the HJB or HJI partial differential equations, design a fault-tolerant controller to meet the requirements of traditional control while ensuring that a certain performance index of the control process reaches the optimal value, so as to minimize the performance index while achieving the control objective. As one of the main methods of modern control theory, the inverse optimal control theory has been applied in many fields such as industrial control and space technology. Therefore, how to utilize the advantages of inverse optimal control in the dynamic system composed of unmanned helicopter formations, combine it with the fault-tolerant control mechanism, and design a fault estimation observer and the corresponding inverse optimal cooperative fault-tolerant control mechanism based on intelligent control technology, so that it can not only simplify the design process of the optimal fault-tolerant controller, but also ensure that the formation system has good robustness and fault tolerance, and achieve the minimization of the performance index, is a difficult problem to be solved in the follow-up of this application. Summary of the Invention

[0006] The embodiment of the present application provides an inverse optimal cooperative fault-tolerant control method for unmanned helicopters based on a composite learning strategy, which can solve the problem that the current unmanned helicopter formation is usually a nonlinear system with underactuation, strong coupling or uncertainty. During the process of the formation executing tasks, various unexpected factors cause components such as actuators and sensors to fail and deviate from the current task, affecting the performance and stability of a single unmanned helicopter within the formation, and making it difficult to achieve the formation cooperative control objective.

[0007] In the first aspect, the embodiment of the present application provides an inverse optimal cooperative fault-tolerant control method for unmanned helicopters based on a composite learning strategy, including the following steps:

[0008] Step 1: Construct an unmanned helicopter formation system composed of a virtual leader and N followers. By establishing the corresponding three-dimensional coordinate system of the system, divide the model of the i-th six-degree-of-freedom unmanned helicopter in the case of actuator failure into a position loop subsystem and an attitude loop subsystem, so as to track the trajectory information provided by the virtual leader in both position and attitude aspects and maintain the preset relative position within the formation.

[0009] Step 2: Use the series-parallel estimation model to approximately estimate the system states of the position loop subsystem and the attitude loop subsystem respectively to obtain predicted values, construct the learning rate of the neural network weights according to the prediction error between the predicted values and the actual values, and adjust the neural network weights according to the learning rate to approximate the actual values of the nonlinear functions in the two subsystems; Based on the composite learning strategy, combine the output of the series-parallel estimation model and the learning results of the neural network, design a fault estimation observer, and obtain the lumped actuator fault information of the subsystem according to the fault estimation observer.

[0010] Step 3: According to the lumped actuator fault information, combined with the inverse optimization theory, and based on the interaction information between adjacent unmanned helicopters in the formation system, design an inverse optimization active fault-tolerant controller for the position loop subsystem and the attitude loop subsystem respectively; compensate for the fault impact by reconstructing the control law to achieve accurate tracking of the trajectory information provided by the virtual leader on the X-axis, Y-axis, Z-axis, and yaw angle, and maintain the consistency of the roll angle and pitch angle; minimize the cost function by optimizing the performance index to ensure the strong robustness and control accuracy of the formation system.

[0011] In a possible implementation manner of the first aspect, the above step 1 includes:

[0012] Construct a communication topology network for the unmanned helicopter formation system, where the virtual leader and the followers communicate with a directed graph, and the followers communicate with an undirected graph, and describe the communication topology relationship between the virtual leader and each follower, and between every two followers according to graph theory knowledge;

[0013] The directed graph is represented by where is the set of followers; is the communication connection set between the followers, (i, j) ∈ ε means that the i-th follower can obtain the information of the j-th follower, and j ∈ N i ={j|(i, j) ∈ ε}; is the weight connection matrix, a ij is the communication weight between the i-th follower and the j-th follower; if a ij =1, it means that the i-th follower communicates with the j-th follower, otherwise a ij =0; if it satisfies a ij >0, a ji >0 and a ij =a ji , then this communication topology is an undirected graph; if there is a root node in the communication topology graph, and there is at least one path from this node to any follower in the connected graph, then this communication topology is a strongly connected graph; describe the communication relationship between the leader and the followers with a diagonal matrix, and this matrix is where b i means that the i-th follower communicates with the virtual leader, otherwise b i =0.

[0014] Optionally, in another possible implementation manner of the first aspect, the above step 1 further includes:

[0015] The model of the i-th six-degree-of-freedom unmanned helicopter under actuator fault is expressed as:

[0016]

[0017] Among them, P i =[x i ,y i ,z i ] T , V i =[u i ,v i ,ω i ] T and Φ i =[φ i ,θ i ,ψ i ] T Respectively represent the position, velocity and Euler angle vector defined in the body coordinate system; Λ i =[p i ,q i ,r i ] T represents the Euler angular velocity defined in the inertial coordinate system; g represents the acceleration due to gravity; e3 = [0,0,1] T ; Z ωi is a constant related to the main rotor speed, blade radius, air density and number of blades; Z coi A constant related to the servo input ratio of the main rotor speed, blade radius and main rotor total pitch angle; N coi is the total coupling coefficient of the main rotor; J i =diag{J xi ,J yi ,J zi} represents the diagonal inertia matrix; δ coi Indicates the main rotor total pitch; denote the total pitch, lateral and longitudinal cyclic pitch of the tail rotor under actuator failure respectively; in addition, the rotation matrix R i , a skew-symmetric matrix Posture kinematics matrix π i And the parameter matrix A i , B i Respectively expressed as:

[0018]

[0019] The parameter τ in the above matrix mi , L bi , L ai , M bi , M ai , N ri , L loi , L lai , M loi , M lai , N coi All represent constants; S (·) , C(·) , T (·) Represent the sine, cosine and tangent values of the corresponding Euler angles respectively; the fault model of the actuator is where θ={coι, loι, laι, taι}, ρ ιm ∈(0,1],m=1,2,3,4和ξ ιm represent partial failure fault factor and bias fault respectively;

[0020] The i-th six-degree-of-freedom unmanned helicopter model is divided into a position loop subsystem and an attitude loop subsystem, where the position loop subsystem is expressed as follows:

[0021]

[0022] Among them, x i ,y i , z i The second derivative of Respectively represent the acceleration in the three directions of X-axis, Y-axis and Z-axis; is the actuator output, and has:

[0023]

[0024] in, as well as are the lumped actuator faults to be estimated;

[0025] The form of the attitude loop subsystem is as follows:

[0026]

[0027] in,

[0028] And there is Θ i2 =L loi ((ρ i2 -1)δ loi +ξ i2 )+L lai ((ρ i3 -1)δ lai +ξ i3 )Θ i3 =M loi ((ρ i2 -1)δ loi +ξ i2 )+M lai ((ρ i3 -1)δ lai +ξ i3 ) and Θ i4 =N tai ((ρ i4 -1)δtai +ξ i4 ) + N coi ((ρ i1 -1)δ coi +ξ i1 ) is the lumped actuator fault to be estimated and compensated.

[0029] Optionally, in another possible implementation of the first aspect, the above step 2 includes:

[0030] The tracking error of the unmanned helicopter formation can be expressed as:

[0031]

[0032] where υ = {φ, θ, ψ}, a in and are the virtual control rates in the two subsystems of the position loop and the attitude loop;

[0033] For the position loop subsystem in step 1, a fault estimation observer based on a composite learning strategy combining a series-parallel estimation model and a neural network is designed, where the subsystem state estimated by the series-parallel estimation model the learning rate of the neural network weights, and the estimated value of the lumped fault in step 1 are respectively expressed as follows:

[0034]

[0035] where γ in , κ in , L in all represent non-negative design parameters; respectively represent the prediction error, the neural network weight estimation error, and the lumped actuator fault estimation error, s in represents the basis function;

[0036] For the attitude loop subsystem in step 1, a fault estimation observer based on a composite learning strategy combining a series-parallel estimation model and a neural network is designed, where the subsystem state estimated by the series-parallel estimation model, the learning rate of the neural network weights, and the fault estimation observer based on the composite learning strategy:

[0037]

[0038] where are all non-negative design parameters; and respectively represent the prediction error, the neural network weight estimation error, and the lumped actuator fault estimation error, Represents a basis function.

[0039] Optionally, in another possible implementation of the first aspect, step 3 above includes:

[0040]

[0041] Wherein, is a parameter to be designed; Ψ in and respectively represent known smooth functions that satisfy ;

[0042] Construct the following Lyapunov functional for the position loop subsystem:

[0043]

[0044] Taking the derivative of it and using Young's inequality, we can obtain:

[0045]

[0046] Wherein, and is a non - negative constant. If and hold, the corresponding inequality holds, where That is, there exists a controller that can make the system actually asymptotically convergent and minimize the performance index:

[0047]

[0048] Wherein, and

[0049] Construct the following Lyapunov functional for the attitude loop subsystem:

[0050]

[0051] Taking the derivative of it and using Young's inequality, we can obtain:

[0052]

[0053] Wherein, If holds, then the corresponding inequality holds.

[0054] Wherein, That is, the controller The performance metric can be minimized:

[0055]

[0056] Among them,

[0057] Beneficial effects: In the technical solution of this application, first unify the six-degree-of-freedom dynamic models of each unmanned helicopter affected by actuator failures, divide them into two subsystems, namely the position loop and the attitude loop, and then design a fault estimation observer based on a composite learning method by combining a series-parallel estimation model with a neural network to respectively and real-time estimate the lumped actuator failures and the subsystem states. For the two subsystems respectively, design an inverse optimization cooperative fault-tolerant control strategy based on mutual information and inverse optimization theory to enable the unmanned helicopter follower to track the position and yaw angle of the virtual leader. Thus, the cooperative fault-tolerant controller designed based on the composite learning and inverse optimization theory has good fault compensation performance and strong robustness, and also minimizes the cost function, meeting the dual requirements of the stability of the real-time unmanned helicopter system and cost reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] To more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0059] Figure 1 is a schematic flowchart of an inverse optimization cooperative fault-tolerant control method for an unmanned helicopter based on a composite learning strategy provided by an embodiment of this application;

[0060] Figure 2 is a communication topology structure diagram of a virtual leader-follower provided by an embodiment of this application;

[0061] Figure 3 is a diagram of the tracking trajectory and deflection angle of each unmanned helicopter follower provided by an embodiment of this application;

[0062] Figure 4 is a formation tracking effect diagram on a three-dimensional plane provided by an embodiment of this application;

[0063] Figure 5 is the lumped fault Θ xi , Θ yi , Θ zi of the actuator of the position loop subsystem and its estimated value diagram;

[0064] Figure 6It is the schematic diagram of the lumped fault Θ of the attitude loop subsystem actuator provided by an embodiment of the present application i1 , Θ i2 , Θ i3 and its estimated value. Specific embodiments

[0065] In the following description, specific details such as specific system structures and technologies are presented for the purpose of illustration rather than limitation, so as to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0066] It should be understood that when used in the specification and claims of the present application, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0067] It should also be understood that the term "and / or" as used in the specification and claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0068] As used in the specification and claims of the present application, the term "if" can be interpreted as "when", "once", "in response to determining", or "in response to detecting" according to the context. Similarly, the phrase "if determined" or "if detected [the described condition or event]" can be interpreted as meaning "once determined", "in response to determining", "once detected [the described condition or event]", or "in response to detecting [the described condition or event]" according to the context.

[0069] In addition, in the description of the specification and claims of the present application, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0070] Reference to "one embodiment" or "some embodiments" etc. described in the specification of this application means that a specific feature, structure or characteristic described in connection with the embodiment is included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments" etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized.

[0071] The following describes in detail the method for inverse optimization collaborative fault-tolerant control of an unmanned helicopter based on a composite learning strategy provided by this application with reference to the accompanying drawings.

[0072] Figure 1 The flowchart of a method for inverse optimization collaborative fault-tolerant control of an unmanned helicopter based on a composite learning strategy provided by an embodiment of this application is shown.

[0073] As Figure 1 shown, the method for inverse optimization collaborative fault-tolerant control of an unmanned helicopter based on a composite learning strategy includes the following steps:

[0074] Step 101: Construct an unmanned helicopter formation system composed of a virtual leader and N followers. By establishing the corresponding three-dimensional coordinate system of the system, the model of the ith six-degree-of-freedom unmanned helicopter in the case of actuator failure is divided into a position loop subsystem and an attitude loop subsystem, so as to track the trajectory information provided by the virtual leader in both position and attitude aspects and maintain the preset relative position inside the formation.

[0075] Furthermore, in the embodiment of this application, the above step 101 includes:

[0076] Construct a communication topology network of the unmanned helicopter formation system, where the virtual leader and the followers communicate with a directed graph, and the followers communicate with an undirected graph among themselves, and describe the communication topology relationship between the virtual leader and each follower, and between every two followers according to graph theory knowledge.

[0077] The directed graph is represented by where is the set of followers; is the communication connection set among the followers. (i, j) ∈ ε means that the ith follower can obtain the information of the jth follower, and j ∈ N i ={j|(i, j) ∈ ε}; is the weight connection matrix, a ijis the communication weight value between the i-th follower and the j-th follower; if a ij = 1, it means that the i-th follower communicates with the j-th follower, otherwise a ij = 0; if a ij > 0, a ji > 0 and a ij = a ji , then the communication topology is an undirected graph; if there is a root node in the communication topology graph, and there is at least one path from this node to any follower in the connected graph, then the communication topology is a strongly connected graph; the communication relationship between the leader and the followers is described by a diagonal matrix, and this matrix is where b i represents that the i-th follower communicates with the virtual leader, otherwise b i = 0.

[0078] Furthermore, in the embodiment of the present application, in the formation control of unmanned helicopters, both the position loop control and the attitude loop control are issues that need to be focused on. In this example, based on the six-degree-of-freedom rigid body model architecture of unmanned helicopters, a fault-tolerant control study is carried out for the unmanned helicopter formation system under actuator faults. It is assumed that the number of agents in the virtual leader structure is N + 1, which are the virtual leader marked as 0, and the unmanned helicopter followers marked as i = 1, 2,..., N. The above step 101 further includes: [[ID=2s]]

[0079] The model of the i-th six-degree-of-freedom unmanned helicopter under actuator faults is expressed as:

[0080]

[0081] where P i = [x i , y i , z i T , V i = [u i , v i , ω i T and Φ i = [φ i , θ i , ψ i T respectively represent the position, velocity, and Euler angle vectors defined in the body coordinate system; Λ i = [p i , q i , r i T represents the Euler angular velocity defined in the inertial coordinate system; g represents the acceleration due to gravity; e3 = [0, 0, 1] T ; Z​​​​ωi is a constant related to the main rotor speed, blade radius, air density and number of blades; Z coi A constant related to the servo input ratio of the main rotor speed, blade radius and main rotor total pitch angle; N coi is the total coupling coefficient of the main rotor; J i =diag{J xi ,J yi ,J zi} represents the diagonal inertia matrix; δ coi Indicates the main rotor total pitch; denote the total pitch, lateral and longitudinal cyclic pitch of the tail rotor under actuator failure respectively; in addition, the rotation matrix R i , a skew-symmetric matrix Posture kinematics matrix π i And the parameter matrix A i , B i Respectively expressed as:

[0082]

[0083] The parameter τ in the above matrix mi , L bi , L ai , M bi , M ai , N ri , L loi , L lai , M loi , M lai , N coi All represent constants; S (·) , C (·) , T (·) Represent the sine, cosine and tangent values of the corresponding Euler angles respectively; the fault model of the actuator is where θ={coι, loι, laι, taι}, ρ ιm ∈(0,1],m=1,2,3,4和ξ ιm represent partial failure fault factor and bias fault respectively;

[0084] The i-th six-degree-of-freedom unmanned helicopter model is divided into a position loop subsystem and an attitude loop subsystem, where the position loop subsystem is expressed as follows:

[0085]

[0086] Among them, x i ,y i , z i The second derivative of Respectively represent the acceleration in the three directions of X-axis, Y-axis and Z-axis; is the actuator output, and has:

[0087]

[0088] in, as well as are the lumped actuator faults to be estimated;

[0089] The form of the attitude loop subsystem is as follows:

[0090]

[0091] in, And there is Θ i2 =L loi ((ρ i2 -1)δ loi +ξ i2 )+L lai ((ρ i3 -1)δ lai +ξ i3 )Θ i3 =M loi ((ρ i2 -1)δ loi +ξ i2 )+M lai ((ρ i3 -1)δ lai +ξ i3 ) and Θ i4 =N tai ((ρ i4 -1)δ tai +ξ i4 )+N coi ((ρ i1 -1)δ coi +ξ i1 ) is the lumped actuator fault to be estimated and compensated.

[0092] It should be noted that in the fault-tolerant control problem of unmanned helicopter formation, based on the two subsystem models of position loop and attitude loop mentioned above, the control goal in this example is to design a fault estimation observer based on a composite learning strategy to achieve the lumped fault Θ xi ,Θ yi ,Θ zi ,Θ i1 ,Θ i2 ,Θ i3 The estimated fault information is used to design an inverse optimization active fault-tolerant control mechanism so that all followers can track the trajectory of the virtual leader and maintain the desired relative position, that is:

[0093]

[0094] in, x0, y0, z0 and ψ0 represent the position and yaw angle of the virtual leader, respectively, is the expected relative distance between the i-th unmanned helicopter follower and the virtual leader.

[0095] Step 102: Use the series-parallel estimation model to approximate the system states of the position loop subsystem and the attitude loop subsystem to obtain predicted values, and construct a learning rate for the neural network weights based on the prediction error between the predicted values and the actual values. Adjust the neural network weights based on the learning rate to approximate the actual values of the nonlinear functions in the two subsystems. Based on a composite learning strategy, combine the output of the series-parallel estimation model with the learning results of the neural network to design a fault estimation observer, and obtain lumped actuator fault information of the subsystems based on the fault estimation observer.

[0096] Furthermore, in the embodiment of the present application, the above step 102 includes:

[0097] In step 1021, the tracking error of the unmanned helicopter formation can be expressed as:

[0098]

[0099] in, υ={φ,θ,ψ}, a in and a in is the virtual control rate in the position loop and attitude loop subsystems;

[0100] Step 1022: For the position loop subsystem in step 101, a fault estimation observer based on a composite learning strategy combining a series-parallel estimation model and a neural network is designed, wherein the subsystem state estimated by the series-parallel estimation model is Learning rate of neural network weights and estimation of lumped fault They are represented as follows:

[0101]

[0102] in, γ in , κ in , L in All represent non-negative design parameters; and Denote prediction error, neural network weight estimation error and lumped actuator fault estimation error, s in represents the basis function;

[0103] Step 1023, a fault estimation observer based on a composite learning strategy that combines a series-parallel estimation model and a neural network is designed for the attitude loop subsystem in Step 101, where the subsystem state estimated by the series-parallel estimation model, the learning rate of the neural network weights, and the fault estimation observer based on the composite learning strategy:

[0104]

[0105] where are all non-negative design parameters; and respectively represent the prediction error, the neural network weight estimation error, and the lumped actuator fault estimation error, represents the basis function.

[0106] As a possible implementation, in an embodiment of the present application, according to the position loop and attitude loop subsystems of the i-th unmanned helicopter follower, a fault estimation observer based on a composite learning strategy is designed to obtain the accurate information of the lumped actuator fault Θ xi , Θ yi , Θ zi , Θ i1 , Θ[[ID=2x6]] i2 , Θ i3 , and the formation tracking error can be expressed as:

[0107]

[0108] where υ = {φ, θ, ψ}, a in and are the virtual control laws to be designed subsequently in the position loop and attitude loop subsystems.

[0109] Taking the derivative of the formation tracking error e in of the position loop subsystem, we can obtain:

[0110] where

[0111] The virtual control law designed for the position loop subsystem then we have where k in is a non-negative design parameter.

[0112] For the non-linear function f in in the position loop subsystem, define fin where L

[0113] By taking the derivative of the error ein Derivation:

[0114]

[0115] Among them, s in and ∈ in respectively represent the neural network weights, basis functions, and estimation errors.

[0116] Next, a series - parallel estimation model is introduced, and a fault estimation observer based on a composite learning strategy is designed in combination with a neural network. The state of the subsystem estimated by the series - parallel estimation model The learning rate of the neural network weights and the estimated value of the lumped fault are respectively expressed as follows:

[0117]

[0118] Among them, γ in , κ in , L in all represent non - negative design parameters, and respectively represent the prediction error, neural network weight estimation error, and lumped actuator fault estimation error.

[0119] Similarly, a neural network is also used to estimate the non - linear function in the attitude subsystem and the tracking error is derived to obtain:

[0120]

[0121] Among them, is a non - negative design parameter, and respectively represent the neural network weights, basis functions, and estimation errors, g i1 = 1, and

[0122] A virtual control law is designed as follows:

[0123]

[0124] Among them, are all non - negative design parameters, and the estimated values of the network weights and the subsystem states are respectively expressed as and is a non - negative design parameter.

[0125] The non - linear function f i1 = q i ri (J zi -J yi ) / J xi -τ i L bi p i -τ i L ai q i ,,f i2 =p i r i (J xi -J zi ) / J yi -τ i M bi p i -τ i M ai q i and f i3 =p i q i (J yi -J xi ) / J zi +N ri r i Neural networks are also used for estimation, and the definition in All of them are non-negative design parameters. The subsystem states estimated by the series-parallel estimation model, the learning rate of the neural network weights, and the fault estimation observer based on the composite learning strategy can be designed as follows:

[0126]

[0127] in are all non-negative design parameters, and denote the prediction error, neural network weight estimation error and lumped actuator fault estimation error, respectively.

[0128] Step 103: Based on the aggregated actuator fault information and inverse optimization theory, inverse optimization active fault-tolerant controllers are designed for the position loop subsystem and attitude loop subsystem respectively, based on the interaction information between adjacent unmanned helicopters in the formation system. The control law is reconstructed to compensate for the impact of the fault, and the trajectory information provided by the virtual leader is accurately tracked on the X-axis, Y-axis, Z-axis, and yaw angle, while maintaining the consistency of the roll angle and pitch angle. The cost function is minimized by optimizing the performance index to ensure the strong robustness and control accuracy of the formation system.

[0129] Furthermore, in the embodiment of the present application, the above step 103 includes:

[0130]

[0131] Among them, is the parameter to be designed; Ψ in and respectively represent known smooth functions that satisfy .

[0132] Construct the following Lyapunov functional for the position loop subsystem:

[0133]

[0134] Take the derivative of it and use Young's inequality to obtain:

[0135]

[0136] Among them, and is a non - negative constant. If and hold, the corresponding inequality holds, where That is, there exists a controller that can make the system actually asymptotically convergent and minimize the performance index:

[0137]

[0138] Among them, and

[0139] Construct the following Lyapunov functional for the attitude loop subsystem:

[0140]

[0141] Take the derivative of it and use Young's inequality to obtain:

[0142]

[0143] Among them, If holds, then the corresponding inequality holds.

[0144] Among them, That is, the controller can minimize the performance index:

[0145]

[0146] Among them,

[0147] The inverse-optimization cooperative fault-tolerant control method for unmanned helicopters based on a composite learning strategy provided by this application first unifies the six-degree-of-freedom dynamic models of each unmanned helicopter affected by actuator faults, divides them into two subsystems: a position loop and an attitude loop, then designs a fault estimation observer based on a composite learning method through a series-parallel estimation model combined with a neural network to respectively and real-time estimate the lumped actuator faults and subsystem states, and respectively designs an inverse-optimization cooperative fault-tolerant control strategy for the two subsystems based on interaction information and inverse-optimization theory to achieve the tracking of the virtual leader's position and yaw angle by the unmanned helicopter follower. Thus, the cooperative fault-tolerant controller designed based on the composite learning and inverse-optimization theory has good fault compensation performance and strong robustness, and also minimizes the cost function, meeting the dual requirements of the stability of the real-time unmanned helicopter system and cost reduction.

[0148] The above embodiments demonstrate the overall process of the inverse-optimization cooperative fault-tolerant control method for unmanned helicopters based on a composite learning strategy proposed by this application. Next, by building a Simulink module in Matlab, a formation system composed of a virtual leader and three unmanned helicopters is used for simulation verification to prove the effectiveness and rationality of the inverse-optimization cooperative fault-tolerant controller for unmanned helicopter formations designed by this application, that is, the expected formation mission can be achieved.

[0149] The structure of the overall formation system and the communication topology structure are as Figure 2 shown, and the communication connection weight values between each agent are 0 or 1.

[0150] The system parameters of the i-th unmanned helicopter are m i = 8.2 Kg, N mi = -0.3705 s -2 , Z ωi = -0.7615 s -1 , Z mi = -131.4125 m / rad·s 2 , g = 9.8 m / s 2 , J i = diag{0.18, 0.34, 0.28}, i = 1, 2, 3, 4, and the matrix parameters are: A i = diag{-48.1757, -25.5048, -0.9808} s -1

[0151]

[0152] Moreover, the reference trajectory and reference yaw angle that each unmanned helicopter within the formation needs to track the virtual leader are respectively \(x_0 = 15\sin(0.2t)\), \(y_0 = 15\cos(0.2t)\), \(z_0 = 0.1t\) and \(\psi_0 = \sin(0.1t)\). The relative distance between formations is and The relevant design parameters of the composite learning fault estimation observer and the inverse optimization cooperative fault-tolerant controller are respectively assumed to be and The actuator faults of the main rotor collective pitch angle, tail rotor collective pitch angle, lateral cyclic pitch angle and longitudinal cyclic pitch angle are respectively assumed that when \(t\geq3s\), the actuator outputs and

[0153] To verify the effect of the inverse optimization cooperative fault-tolerant control method for the unmanned helicopter of the present invention, the following will conduct simulation verification by building the Simulink module in Matlab. In the initial state \(P_1 = [0,0,0]\) T , \(P_2 = [0,1,0]\) T , \(P_3 = [1,0,0]\) T , Figures 3 - 6 shows the performance of the inverse optimization cooperative fault-tolerant controller designed in this example. Figure 3 are the tracking trajectories of each unmanned helicopter follower on the x-axis, y-axis, z-axis and deflection angle, Figure 4 is the three-dimensional plane formation effect diagram. From this Figures 3 - 4 it can be seen that even affected by sudden faults, the inverse optimization cooperative fault-tolerant controller in this example can still enable the formation system composed of multiple unmanned helicopters to form the desired formation and maintain stability according to requirements by reconstructing the controller. Figure 5 and Figure 6 respectively show the estimation of the lumped fault \(\Theta\) of the actuator xi , \(\Theta\) yi , \(\Theta\) zi and \(\Theta\) i1 , \(\Theta\) i2 , \(\Theta\) i3 and its estimated value. The designed fault estimation observer can achieve accurate estimation of the lumped fault.

[0154] It should be understood that the magnitude of the serial numbers of the steps in the above embodiments does not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0155] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included within the protection scope of the present application.

Claims

1. An inverse optimization cooperative fault-tolerant control method for an unmanned helicopter based on a composite learning strategy, characterized in that Specifically, it includes the following steps: Step 1: Construct a formation system of unmanned helicopters consisting of a virtual leader and N followers. By establishing the corresponding three-dimensional coordinate system of the system, divide the model of the i-th six-degree-of-freedom unmanned helicopter in the case of actuator failure into a position-loop subsystem and an attitude-loop subsystem, so as to track the trajectory information provided by the virtual leader in both position and attitude aspects and maintain the preset relative positions within the formation; Step 2: Use the series-parallel estimation model to approximately estimate the system states of the position-loop subsystem and the attitude-loop subsystem respectively to obtain predicted values, and construct the learning rate of the neural network weights according to the prediction error between the predicted values and the actual values, and adjust the neural network weights according to the learning rate to approximate the actual values of the nonlinear functions in the two subsystems; Based on the composite learning strategy, combine the output of the series-parallel estimation model and the learning result of the neural network, design a fault estimation observer, and obtain the lumped actuator fault information of the subsystem according to the fault estimation observer; Step 3: According to the lumped actuator fault information, combined with the inverse optimization theory, based on the interaction information between adjacent unmanned helicopters in the formation system, design inverse optimization active fault-tolerant controllers for the position-loop subsystem and the attitude-loop subsystem respectively; compensate for the fault impact by reconstructing the control law, and achieve accurate tracking of the trajectory information provided by the virtual leader on the X-axis, Y-axis, Z-axis and yaw angle, and maintain the consistency of the roll angle and pitch angle; Minimize the cost function by optimizing the performance index to ensure the strong robustness and control accuracy of the formation system.

2. The inverse optimization cooperative fault-tolerant control method for an unmanned helicopter based on a composite learning strategy according to claim 1, characterized in that The said Step 1 includes: Construct the communication topology network of the unmanned helicopter formation system, where the virtual leader communicates with the followers through a directed graph, and the followers communicate with each other through an undirected graph, and describe the communication topology relationship between the virtual leader and each follower, and between every two followers according to graph theory knowledge; The directed graph is represented by where is the set of followers; is the set of communication connections between followers. (i, j) ∈ ε means that the i-th follower can obtain the information of the j-th follower, and j ∈ N i = {j|(i, j) ∈ ε}; is the weight connection matrix, and a ij is the communication weight between the i-th follower and the j-th follower; if a ij = 1, it means that the i-th follower communicates with the j-th follower, otherwise a ij = 0; if a ij > 0, a ji > 0 and a ij = a ji , then the communication topology is an undirected graph; if there is a root node in the communication topology graph and there is at least one path from this node to any follower in the connected graph, then the communication topology is a strongly connected graph; the communication relationship between the leader and the followers is described by a diagonal matrix, and this matrix is where b i means that the i-th follower communicates with the virtual leader, otherwise b i = 0.

3. The inverse optimization cooperative fault-tolerant control method for an unmanned helicopter based on a composite learning strategy according to claim 1, wherein The said Step 1 also includes: The model of the i-th six-degree-of-freedom unmanned helicopter under actuator failure is expressed as: Among them, P i = [x i , y i , z i T , V i = [u i , v i , ω i T and Φ i = [φ i , θ i , ψ i T respectively represent the position, velocity, and Euler angle vector defined in the body coordinate system; Λ i = [p i , q i , r i T represents the Euler angular velocity defined in the inertial coordinate system; g represents the gravitational acceleration; e3 = [0, 0, 1] T ; Z ωi is a constant related to the main rotor speed, blade radius, air density, and number of blades; Z coi is a constant related to the servo input ratio of the main rotor speed, blade radius, and total pitch angle of the main rotor; N coi is the total coupling coefficient of the main rotor; J i = diag{J xi , J yi , J zi} represents the diagonal inertia matrix; δ coi represents the total pitch of the main rotor; respectively represent the total pitch of the tail rotor, lateral and longitudinal cyclic pitches under actuator faults; In addition, the rotation matrix R i , the skew-symmetric matrix the attitude kinematics matrix Π i and the parameter matrices A i , B i are respectively expressed as:​​​​ The parameter τ in the above matrix mi , L bi , L ai , M bi , M ai , N ri , L loi , L lai , M loi , M lai , N coi all represent constants; S (·) , C (·) , T (·) respectively represent the sine value, cosine value and tangent value of the corresponding Euler angle; the fault model of the actuator is where ρ ιm ∈(0,1], m = 1, 2, 3, 4 and ξ ιm respectively represent the partial failure fault factor and the bias fault; Divide the model of the i-th six-degree-of-freedom unmanned helicopter into a position-loop subsystem and an attitude-loop subsystem, where the position-loop subsystem is expressed as follows: where x i , y i , z i The second derivatives of represent the accelerations in the three directions corresponding to the X-axis, Y-axis, and Z-axis respectively; is the output of the actuator, and there is: wherein, and are respectively the lumped actuator faults to be estimated; The form of the attitude-loop subsystem is as follows: Among them, And there is Θ i2 = L loi ((ρ i2 - 1)δ loi + ξ i2 ) + L lai ((ρ i3 - 1)δ lai + ξ i3 )Θ i3 = M loi ((ρ i2 - 1)δ loi + ξ i2 ) + M lai ((ρ i3 - 1)δ lai + ξ i3 ) and Θ i4 = N tai ((ρ i4 - 1)δ tai + ξ i4 ) + N coi ((ρ i1 - 1)δ coi + ξ i1 ) are the lumped actuator faults to be estimated and compensated.

4. The method for inverse optimization cooperative fault-tolerant control of an unmanned helicopter based on a composite learning strategy according to claim 3, wherein The said Step 2 includes: The tracking error of the unmanned helicopter formation can be expressed as: Among them, υ = {φ, θ, ψ}, a in and are the virtual control rates in the two subsystems of the position loop and the attitude loop; A fault estimation observer based on a composite learning strategy that combines a series-parallel estimation model and a neural network is designed for the position loop subsystem in Step 1, where the subsystem state estimated by the series-parallel estimation model the learning rate of the neural network weights and the estimated value of the lumped fault in Step 1 are respectively expressed as follows: wherein, γ in , κ in , L in all represent non - negative design parameters; and respectively represent the prediction error, the neural network weight estimation error and the lumped actuator fault estimation error, s in represents the basis function; Design a fault estimation observer based on the composite learning strategy that combines the series-parallel estimation model and the neural network for the attitude-loop subsystem in Step 1, where the subsystem state estimated by the series-parallel estimation model, the learning rate of the neural network weights, and the fault estimation observer based on the composite learning strategy: wherein are all non - negative design parameters; and respectively represent the prediction error, the neural network weight estimation error, and the lumped actuator fault estimation error, represents the basis function.

5. The method for inverse optimization collaborative fault-tolerant control of an unmanned helicopter based on a composite learning strategy according to claim 4, wherein The said Step 3 includes: Among them, is the parameter to be designed; Ψ in and respectively represent known smooth functions that satisfy ; Construct the following Lyapunov functional for the position-loop subsystem: Derive it and use the Young's inequality to obtain: Among them, and is a non - negative constant. If and hold, there will be an inequality holding accordingly. Among them That is, under a fault, the controller can make the system actually asymptotically converge and minimize the performance index: Among them, and Construct the following Lyapunov functional for the attitude-loop subsystem: Derive it and use the Young's inequality to obtain: Among them, If holds, then the corresponding inequality holds. Among them, i.e., the controller can minimize the performance index: Among them,

Images