CN115356929A - Proportional allowable tracking control method for actuator attack singularity multi-agent system - Google Patents
Proportional allowable tracking control method for actuator attack singularity multi-agent system Download PDFInfo
- Publication number
- CN115356929A CN115356929A CN202211012274.3A CN202211012274A CN115356929A CN 115356929 A CN115356929 A CN 115356929A CN 202211012274 A CN202211012274 A CN 202211012274A CN 115356929 A CN115356929 A CN 115356929A
- Authority
- CN
- China
- Prior art keywords
- agent system
- singular
- matrix
- follower
- agent
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a proportional allowable tracking control method for a strange multi-agent system attacked by an executor, which comprises the following steps: step 1: constructing a follower singular multi-agent system, a leader singular multi-agent system and a nonlinear actuator attack model containing time lag; performing state estimation on the attack of the nonlinear actuator by adopting a radial neural network; step 2, designing a state observer for each follower agent; and step 3: constructing a self-adaptive distributed controller and an integral sliding-mode surface equation containing a singular matrix; and 4, step 4: analyzing the tolerance of a closed-loop system under the action of an adaptive controller through a singular system tolerance theory and a robust stability theory, and providing a singular multi-agent system proportion tolerance following control method; and 5: and (4) verifying the limited time accessibility of the designed integral sliding mode surface equation. The invention realizes the proportional allowable tracking control of the singular multi-agent system under the attack of an unknown nonlinear actuator, so that the system realizes the tracking and has good inhibiting effect on the attack of the actuator.
Description
Technical Field
The invention belongs to the technical field of multi-agent system control, and relates to a strange multi-agent system proportion allowable tracking control method with actuator attack.
Background
In recent years, the problem of distributed consistent control of multi-agent systems has received widespread attention and use. In general, consistent controls can be divided into leader-less consistent controls and leader-tracking controls. The tracking control is that all followers track one or a group of leaders under the action of the distributed controller, so that the method has the advantages of improving the communication efficiency and reducing the communication cost. In practical applications, different agents may perform different tasks. The proportional consistent control of a multi-agent system requires that agents converge according to different proportions, and the final convergence value of an agent is independent of the initial state of the system. Therefore, the proportional consistent control can effectively solve the problem of multi-scale coordination control among the intelligent agents. In some actual complex system modeling, a differential equation and an algebraic constraint equation are combined to establish an accurate system model. The dynamical model with algebraic constraints is called a singular system. A system formed by connecting a plurality of intelligent agents with singular dynamics models through a wired or wireless network is called a singular multi-intelligent-agent system.
The research of the fanciful multi-agent system has mostly focused on the problem of allowable consistent control without the leader, and the problem of allowable tracking control with the leader agent has not been fully studied. Meanwhile, the convergence targets of the singular multi-agent system depend on the initial state of the system, and the ratio-allowed coordination control of the singular multi-agent system with the convergence value independent of the initial state of the system is not involved. During the operation of a multi-agent system, attacks are one of the main threats to the security performance of the system. In order to ensure the security of a multi-agent system under attack conditions, the following two methods are mainly adopted at present. First, an attacked agent is detected, identified and deleted. Secondly, the external attack is restrained by combining the distributed controller under the condition that the attacked agent is not removed, and the system is guaranteed to have good resistance and recovery performance to the external attack. At present, the distributed consistent control research results with actuator attacks are concentrated on a normal multi-agent system, and the problem of allowable tracking control of the singular multi-agent system under the condition of an actuator is not considered yet. The sliding mode control is a nonlinear variable structure control, and when the system state passes through different areas, the sliding mode feedback control structure is controlled according to different rules. The sliding mode control has strong robustness on external interference of the system, can solve the problem of distributed coordination control of the singular multi-agent system with the external interference, and effectively reduces the damage of the external interference on the system performance.
At present, the multi-agent system tracking control research with actuator attack is mostly concentrated on a normal multi-agent system, and a singular multi-agent system with a kinetic model containing algebraic constraints is not fully considered. Further, the scale-tolerant tracking control can effectively solve the problem of multi-scale coordination control among singular multi-agents. Therefore, scale-tolerant tracking control of the singular multi-agent system with an actuator attack is an urgent problem to be solved.
Disclosure of Invention
The invention aims to provide a proportion-allowed tracking control method of a singular multi-agent system with actuator attack, which can provide a self-adaptive distributed controller based on a neural network and sliding mode control to drive the singular multi-agent system to realize the proportion-allowed tracking control.
The technical scheme adopted by the invention is as follows:
the proportional allowable tracking control method for the actuator attacking singular multi-agent system comprises the following steps:
and 5, verifying the limited time accessibility of the designed integral sliding mode surface equation.
The invention is also characterized in that:
the model of the follower singular multi-agent system in the step 1 is specifically as follows:
consider N follower singular multi-agent systems containing nonlinear effector attacks, where the model of the ith follower singular multi-agent system can be represented as:
x i (t)=η i (t),t∈[-h,0] (1)
wherein x i (t)∈R n Is the state of the ith agent, and h (t) is 0-h (t) -h<Infinity andtime-varying time lag of u ir (t)∈R p Is a control input, w i (t)∈R q Is an external disturbance; function eta i (t) is an initial value of agent i. The matrix pair (E, A) is regular; wherein A ∈ R n×n ,B u ∈R n×p And B w ∈R n×q Is a known system parameter matrix, matrix B u Column full rank;
the model of the leader's singular multi-agent system is:
x 0 (t)=η 0 (t),t∈[-h,0] (2)
wherein x 0 (t)∈R n Is the state of the leader's singular multi-agent system. Initial state of leader singular multi-agent system is eta 0 (t);
The nonlinear actuator attack model containing time lag and the estimation method are as follows:
the actuator in the operation of the singular multi-agent system can be attacked, and the safety performance of the system is threatened: assuming that the actuators of each follower's singular multi-agent system are under attackIs different in strength, defines an attack coefficient alpha i ∈[0,1]. Then, the actuator attack function of the follower singular multi-agent system i can be expressed as:
u ir (t)=u i (t)+α i φ i (x i (t),x i (t-h(t)),t) (3)
wherein the content of the first and second substances,
using radial neural networksTo phi i (x i (t),x i (t-h (t)), t) estimating with an estimation error ofWherein the vectorIs the input variable of the neural network with an input deviation of-1, vectorError vectorFor any e i >0 satisfiesMatrix arrayRepresenting the weight of the hidden layer to the output layer,representing the weights of the input layer to the hidden layer. l i The number of nodes is hidden in the neural network, and p is the number of nodes in the output layer. Nonlinear equation σ i (. Is) an input-to-hidden-layer transfer function that can be expressed as a function vector of:wherein
For i =1,2.., N, the hypothetical matrixAndare respectively gamma i And Θ i The estimation matrix of (2); then, the estimation error of the weight matrix can be expressed asAndassuming a non-linear function phi i (x i (t),x i (t-h (t)), t) an estimation function ofThen, the error function:
wherein, the first and the second end of the pipe are connected with each other,is a Jacobian matrix, and the infinitesimal term o (-) is expressed asNorm of residual termThe following conditions are satisfied:wherein beta is i ∈R 4 Is the unknown vector of the vector,is beta i And the corresponding estimation error isFor an arbitrary matrix M ∈ R p×q ,Wherein λ max (M T M) a representation matrix M T The maximum eigenvalue of M;
the step 2 specifically comprises the following steps:
step 2.1, designing a state observer for each follower singular multi-agent system:
whereinIs state estimation, design u ai (t) for reducing non-linear actuator attacks a containing time lag i φ i (x i (t),x i (t-h (t)), t); and is provided withIs an initial value of the observer system;
step 2.2. Defining state errors of follower singular multi-agent system and state observerThe expression for solving the state error equation by equation (8) is:
step 2.3. Construction of proportional errorAnd solving for the proportional errorThe kinetic equation of (a):
wherein the content of the first and second substances,v i and v j Is a proportional function allowing proportional tracking control, and v ij =v i /v j ,i=1,2,...,N,j=0,1,2,...,N。
The step 3 specifically comprises:
step 3.1, designing an integral sliding mode surface equation:
wherein X ∈ R n×n Is an unknown nonsingular matrix, K ∈ R p×n Is the feedback gain matrix to be designed, s is the integral variable, a ij For the connection weight between the follower singular multi-agent system j and the follower singular multi-agent system i, if the follower singular multi-agent system j is communicated with the follower singular multi-agent system i, a ij =a ji >0; otherwise, a ij =0;b i For the connection weights between the follower singular multi-agent and the leader singular multi-agent, if the follower singular multi-agent system i is in communication with the leader singular multi-agent system, b i >0; otherwise, b i =0;
Step 3.2, according to the sliding mode control theory, when the proportion error system reaches the sliding mode surface,andif true; obtaining an equivalent controller u by taking the integral sliding mode surface derivative as zero eqi (t) is:
substituting the equivalent controller into a proportional error equation to obtain a sliding-mode kinetic equation:
and 3.3, designing a self-adaptive distributed controller based on a neural network:
wherein alpha is i Is the attack coefficient, σ, of the actuator under attack i (. Is the input layer to hidden layer transfer function, p i (t)>0 is the neural network based adaptive distributed controller parameter to be solved,wherein sgn (.) is a sign function, for any function x, when x>At 0, sgn (x) =1; sgn (x) =0 when x =0; when x is<At 0, sgn (x) = -1;
the updating rule of the parameters is as follows:
wherein the matrixAndweight matrix gamma from hidden layer to output layer i And the weight Θ of the input layer to the hidden layer i Is determined by the estimation matrix of (a),is an unknown vector beta i Is determined by the estimated value of (c),is a Jacobian matrix, M i1 >0,M i2 >0,M i3 >0 is the gain matrix, ρ i (t)>0 is the adaptive distributed controller parameter, vector based on neural network to be solved
The step 4 specifically comprises the following steps:
And Q 2 ∈R n×n >0, satisfying:
wherein
The adaptive distributed controller (14) based on neural network can drive the state variables of the follower singular multi-agent system (1) to track the state of the leader singular multi-agent system (2) in a fixed proportion, i.e.,establishing all the singular multi-agent systems;
step 4.2 when w i When (t) ≠ 0, given γ>0, if there is a matrixH 1 ∈R n×n And positive definite moment of symmetry Q 1 ∈R n×n ,H 1 ∈R n×n ,H 2 ∈R q×q Satisfy the requirement of
Wherein
Matrix U satisfiesIs a diagonal matrix and feeds back a gain matrixThe adaptive distributed controller (14) based on the neural network can drive the state variables of the follower singular multi-agent system (1) to track the state of the leader singular multi-agent system (2) according to a fixed proportion, and has a restraining effect on the external interference of the follower singular multi-agent system, namely,the external interference of the follower singular multi-agent system is satisfied when all the singular multi-agent systems are established:
the specific method of the step 4.1 is as follows:
step 4.1.1, adopting a singular system model decomposition method to carry out pair inequality A T X+X T A+Q 1 <0 is decomposed to obtain:
thus, det (A) 22 ) Not equal to 0,det (.) represents the determinant of the matrix, i.e. (E, A) no pulse;
step 4.1.2, constructing Lyapunov function of the formula (18)
Wherein, the first and the second end of the pipe are connected with each other,
the Lyapunov function is derived along an error system under the action of a self-adaptive controller based on a neural network, and the following can be obtained:
wherein the content of the first and second substances, matrix of l ij =-a ij All i is not equal to j; according to the formula (16), theThus, error functionAndthe average index is stable; the proportional error tracking system has no pulse and satisfiesThat is, the singular multi-agent system containing time-lapse nonlinear actuator attacks implements scale-tolerant tracking control.
The step 4.2 is specifically as follows:
when w (t) ≠ 0, the Lyapunov function constructed in the equation (18) is combined with the following energy function
The above equation is solved to obtain
Wherein
By means of matricesAndright and left squaring respectively 5×5 Let us orderAnddue to the matrix M L And M R If the matrix is a full rank matrix, J is less than or equal to 0; and feedback the gain matrix
The step 5 specifically comprises the following steps:
and 5.1, if the matrix K meets the condition in the step 4.2 and parameters in the adaptive controller (14) based on the neural network meet:
wherein, for all i =1,.., N,is a normal number; the integral sliding mode surface equation (11) can be reached in limited time;
step 5.2, constructing a Lyapunov function of the formula (22):
the following are obtained by calculation:
wherein the content of the first and second substances,thus, there is a momentSo that all T ≧ T satisfy V s (t) =0 and s (t) =0. Namely, the integral sliding mode surface equation can be reached within a limited time.
The invention has the beneficial effects that:
1. compared with the time-lag independent nonlinear actuator attack, the state and time-lag dependent actuator attack function designed by the invention can better simulate the actuator attack model in the actual system operation. And the invention adopts an estimation model based on a neural network method, thereby realizing the rapid estimation of the nonlinear actuator attack of the singular multi-agent system.
2. The invention designs a novel distributed self-adaptive controller based on sliding mode control and a neural network, the controller effectively reduces the influence of external interference on the tracking performance, and can drive the follower agent state to track the leader agent state according to a fixed proportion;
3. the proportion allowable tracking control of the singular multi-agent system under the condition of the actuator attack comprises the general conditions of allowable tracking control, proportion tracking control of a normal multi-agent system, tracking control of the normal multi-agent system and the like.
Drawings
FIG. 1 is a system control flow diagram of the method of the present invention;
fig. 2 is a communication topology diagram of an agent according to embodiment 1 and embodiment 2 of the present invention;
fig. 3 shows the locus diagrams of the proportional tracking error of the embodiment 1 when i =1,2,3,4 in (a), (b), (c) and (d), respectively;
fig. 4 is a schematic diagram of errors between the actuator attack and the neural network estimation function in the embodiment 1 when i =1,2,3,4 in (a), (b), (c), and (d), respectively;
in fig. 5, (a), (b), (c), and (d) are trajectory diagrams of sliding mode surface equations of example 1 when i =1,2,3,4, respectively;
fig. 6 is a trace diagram of proportional tracking errors of example 2 of the present invention when i =1,2,3,4 in (a), (b), (c), and (d), respectively;
fig. 7 is a schematic diagram of errors between the actuator attack and the neural network estimation function of embodiment 2 of the present invention when i =1,2,3,4 in (a) (b) (c) (d), respectively;
in fig. 8, (a), (b), (c), and (d) are trajectory diagrams of sliding mode surface equations for embodiment 2 of the present invention when i =1,2,3,4, respectively.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention relates to a proportional allowable tracking control method for a singular multi-agent system under attack of an executor, as shown in figure 1, comprising the following steps of:
and 5, verifying the limited time accessibility of the designed integral sliding mode surface equation.
The specific steps of the step 1 are as follows:
step 1.1. Model of follower singular multi-agent system:
consider N follower singular multi-agent systems containing nonlinear actuator attacks, where the model of the ith follower singular multi-agent system can be represented as:
x i (t)=η i (t),t∈[-h,0] (1)
wherein x i (t)∈R n Is the state of the ith agent, and h (t) is 0-h (t) -h<Infinity andtime-varying time lag of u ir (t)∈R p Is a control input, w i (t)∈R q Is an external disturbance. Function eta i (t) is the initial value of agent i. The matrix pair (E, A) is regular. Wherein A ∈ R n×n ,B u ∈R n×p And B w ∈R n×q Is a known system parameter matrix, matrix B u Column full rank.
Step 1.2. The model of the leader's singular multi-agent system is:
x 0 (t)=η 0 (t),t∈[-h,0] (2)
wherein x 0 (t)∈R n Is the state of the leader's singular multi-agent system. Initial state of leader singular multi-agent system is eta 0 (t)。
Step 1.3, a nonlinear actuator attack model containing time lag and estimation:
the actuators in the operation of the singular multi-agent system can be attacked, and the safety performance of the system is threatened. Assuming that the attack strength of actuators of each follower singular multi-agent system is different, defining attack coefficient alpha i ∈[0,1]. Then, the actuator attack function of the follower singular multi-agent system i can be expressed as:
u ir (t)=u i (t)+α i φ i (x i (t),x i (t-h(t)),t) (3)
wherein, the first and the second end of the pipe are connected with each other,
using radial neural networksTo phi i (x i (t),x i (t-h (t)), t) estimating with an estimation error ofWherein the vectorIs an input variable of a neural network with an input offset of-1, vectorError vectorFor any e i >0 satisfiesMatrix ofRepresenting the weight of the hidden layer to the output layer,representing the weights of the input layer to the hidden layer. l. the i The number of nodes is hidden in the neural network, and p is the number of nodes in the output layer. Nonlinear equation σ i (. Is) an input-to-hidden-layer transfer function that can be expressed as a function vector of:wherein
For i =1,2.., N, the hypothetical matrixAndare respectively gamma i And Θ i The estimation matrix of (2). Then, the estimation error of the weight matrix can be expressed asAndassuming a non-linear function phi i (x i (t),x i (t-h(t)),t)Is an estimation function ofThen, the error function:
wherein the content of the first and second substances,is a Jacobian matrix with an infinitesimal term o (-) expressed asNorm of residual termThe following conditions are satisfied:wherein beta is i ∈R 4 Is the unknown vector of the vector,is beta i And the corresponding estimation error isFor an arbitrary matrix M ∈ R p×q ,Wherein λ max (M T M) a representation matrix M T The maximum eigenvalue of M;
wherein the step 2 specifically comprises:
step 2.1, designing a state observer for each follower singular multi-agent system:
whereinIs state estimation, design u ai (t) for reducing non-linear actuator attacks a containing time lag i φ i (x i (t),x i (t-h (t)), t). And isIs an initial value of the observer system.
Step 2.2. Defining state errors of follower singular multi-agent system and state observerThe expression for solving the state error equation by equation (8) is:
step 2.3. Construction of proportional errorAnd solving for the proportional errorThe kinetic equation of (c):
wherein the content of the first and second substances,v i and v j A proportional function for the allowable proportional tracking control, and v ij =v i /v j ,i=1,2,...,N,j=0,1,2,...,N.
Wherein the step 3 specifically comprises:
step 3.1, designing an integral sliding mode surface equation:
wherein X ∈ R n×n Is an unknown nonsingular matrix, K ∈ R p×n Is the feedback gain matrix to be designed, s is the integral variable, a ij For the connection weight between the follower singular multi-agent system j and the follower singular multi-agent system i, if the follower singular multi-agent system j is communicated with the follower singular multi-agent system i, then a ij =a ji >0; otherwise, a ij =0;b i For the connection weights between the follower singular multi-agent and the leader singular multi-agent, if the follower singular multi-agent system i is in communication with the leader singular multi-agent system, b i >0; otherwise, b i And =0. Step 3.2, according to the sliding mode control theory, when the proportion error system reaches the sliding mode surface,andthis is true. Obtaining an equivalent controller u by taking the integral sliding mode surface derivative as zero eqi (t) is:
substituting the equivalent controller into a proportional error equation to obtain a sliding-mode kinetic equation:
step 3.3, designing the self-adaptive distributed controller based on the neural network
Wherein sgn (·) is a sign function, and for an arbitrary function x, sgn (x) =1 when x > 0; sgn (x) =0 when x =0; sgn (x) = -1 when x < 0.
The updating rule of the parameters is as follows:
wherein M is i1 >0,M i2 >0,M i3 >0 is the gain matrix, ρ i (t)>0 is the adaptive distributed controller parameter, vector based on neural network to be solved
Wherein the step 4 specifically comprises:
whereinThe adaptive distributed controller (14) based on neural network can drive the state variables of the follower singular multi-agent system (1) to track the state of the leader singular multi-agent system (2) in a fixed proportion, i.e.,the method is established for all the singular multi-agent systems, and specifically comprises the following steps:
step 4.1.1, adopting a singular system standard model decomposition method to carry out inequality A T X+X T A+Q 1 <0 is decomposed to obtain
Thus, det (A) 22 ) Not equal to 0,det (.) represents the determinant of the matrix, i.e. (E, A) no pulse;
step 4.1.2. Constructing the Lyapunov function of the formula (18):
wherein, the first and the second end of the pipe are connected with each other,
the Lyapunov function is derived along an error system under the action of a self-adaptive controller based on a neural network, and the following can be obtained:
wherein, the first and the second end of the pipe are connected with each other, matrix array l ij =-a ij For all i ≠ j. From the formula (16), it can be obtainedThus, error functionAndthe average index is stable. Error in proportionThe difference tracking system has no pulse and satisfiesThat is, the singular multi-agent system containing time-lapse nonlinear actuator attacks implements scale-tolerant tracking control.
Step 4.2 when w i When (t) ≠ 0, given γ>0, if a matrix existsH 1 ∈R n×n Positive definite moment of symmetry Q 1 ∈R n×n ,H 1 ∈R n×n ,H 2 ∈R q×q Satisfy the requirement of
Wherein
Matrix U satisfiesIs a diagonal matrix and feeds back a gain matrixThe adaptive distributed controller (14) based on the neural network can drive the state variables of the follower singular multi-agent system (1) to track the state of the leader singular multi-agent system (2) according to a fixed proportion, and has a restraining effect on the external interference of the follower singular multi-agent system, namely,the external interference of the follower singular multi-agent system is satisfied when all the singular multi-agent systems are established:
when w (t) ≠ 0, the Lyapunov function constructed in equation (18) is combined with the following energy function:
solving the above equation can obtain:
wherein
By means of matricesAndright and left multiplier xi, respectively 5×5 Let us orderAnddue to the matrix M L And M R All are full rank matrices, then J ≦ 0 may be deduced. And feedback the gain matrix
And 5, verifying the limited time accessibility of the integral sliding mode surface:
step 5.1. If the matrix K satisfies the condition in step 4.2 and the parameters in the adaptive controller (14) based on the neural network satisfy
Wherein, for all i =1,.., N,is a normal number. The integral sliding-mode surface equation (11) is reachable in a finite time.
Step 5.2, constructing a Lyapunov function of the formula (22):
is obtained by calculation
Wherein the content of the first and second substances,thus, there is a momentSo that all T ≧ T satisfy V s (t) =0 and s (t) =0. Namely, the integral sliding mode surface equation can be reached within a limited time.
Example 1
In order to verify the proposed ratio allowable tracking control effect, the method provided by the invention is adopted to carry out simulation verification by using matlab. In this embodiment, FIG. 2 is an experimental topology diagram, which includes a leader node 0 and 4 follower agent nodes 1,2,3,4. The matrix parameters of the agent are
E=[1 0 0;0 1 0;0 0 0],A h =[0.4 -0.3 0.2;0.2 0.4 0.4;0 0 0.1],
A=[1 2 1;3 2 4;1 2 1],B u =[1 2 0] T ,B w =[0 1 -1] T
The time-lag nonlinear actuator attack function is defined asAnd let h (t) =0.2sin (t). Assuming that the attack coefficient of each follower singular multi-agent system i is alpha i =1, i.e. all follower singular multi agent systems are exposedAnd (4) attack of the section. The scale function of the allowable tracking is v 0 =1,v 1 =1.5,v 2 =1,v 3 =1.5,v 4 And =1. The simulation results of FIG. 3 show the proportional allowable tracking error e i (t)=x i (t)-v i0 x 0 (t) state trace. From FIG. 3, the state trace e of the scale-tolerant tracking error can be seen i (t)=x i (t)-v i0 x 0 (t) converges to zero after 6s, i.e. proportional tolerant tracking control can be achieved. Suppose the number of hidden layers of the neural network is 10, i.e. node =10, and the initial value of the weight function isAndand matrixM i2 =4diag(1,1,1,1),M i3 =5diag (1,1,1,1). FIG. 4 is a trace of error between time-lapse nonlinear actuator attacks and neural network estimated parametersAs can be seen from fig. 4, the designed neural network function can realize fast estimation of the time-lag nonlinear actuator attack of the follower singular multi-agent system i =1,2,3,4. Fig. 5 is a track of a sliding mode surface equation, and it can be known from the figure that the sliding mode surface equation is asymptotically stable.
Example 2
Considering the singular multi-agent system scale tolerant tracking control without time lag, i.e. the time lag function h (t) =0. Wherein the matrix parameters of the singular multi-agent system are the same as those in embodiment 1, and the actuator attack model isThe scale-tolerant tracking function and neural network are the same as in example 1, and the singular agent system i assumes an attack coefficient α for each follower i =1, i.e. all follower singular multi-agent systems are subject to external attacks. The scale function of the allowable tracking is v 0 =1,v 1 =1.5,v 2 =1,v 3 =1.5,v 4 =1, FIG. 6 is the singular multi-agent system proportional tracking error e without time lag i (t)=x i (t)-v i0 x 0 The state trajectory of (t) is faster in the rate of convergence of the proportional tracking error function of the singular multi-agent system containing no time lag than in example 1. Further, fig. 7 is a difference value between the neural network estimation function and the actuator attack function without time lag, and it can be known from the figure that the designed neural network function can realize fast estimation of the actuator attack without time lag. Fig. 8 is a track of a sliding mode surface equation without time lag, and it can be known from the figure that the designed integral sliding mode surface equation is asymptotically stable.
Claims (6)
1. The proportional allowable tracking control method for the actuator attack singularity multi-agent system is characterized by comprising the following steps of:
step 1, constructing a follower singular multi-agent system, a leader singular multi-agent system and a nonlinear actuator attack model containing time lag; performing state estimation on the attack of the nonlinear actuator by adopting a radial neural network;
step 2, designing a state observer for each follower intelligent agent and constructing a state error and a proportion error;
step 3, constructing an integral sliding mode surface equation containing a singular matrix and a self-adaptive distributed controller based on a neural network;
step 4, analyzing the tolerance of the closed-loop system under the action of the self-adaptive controller through a singular system tolerance theory and a robust stability theory, and providing a singular multi-agent system proportion tolerance following control method;
step 5, verifying the limited time accessibility of the designed integral sliding mode surface equation;
the step 2 specifically comprises:
step 2.1, designing a state observer for each follower singular multi-agent system:
wherein E, A h ,B u For the parameter matrix of the known singular multi-agent system, h (t) is equal to or more than 0 and equal to or less than h (t)<Infinity andtime-varying time lag of u i (t) is the distributed controller to be designed, is state estimation, design u ai (t) for reducing non-linear actuator attacks a containing time lag i φ i (x i (t),x i (t-h (t)), t); and is provided withIs an initial value of the observer system;
step 2.2, state errors of follower singular multi-agent system and state observer are definedWherein x is i (t) is the state of the ith agent, and the expression for solving the state error equation by equation (8) is:
step 2.3, construction of proportional errorAnd solving for the proportional errorThe kinetic equation of (a):
wherein the content of the first and second substances,v i and v j A proportional function for the allowable proportional tracking control, and v ij =v i /v j ,i=1,2,...,N,j=0,1,2,...,N;
The step 3 specifically comprises:
step 3.1, designing an integral sliding mode surface equation:
where X is an unknown nonsingular matrix, K is a feedback gain matrix to be designed, s is an integral variable, a ij For the connection weight between the follower singular multi-agent system j and the follower singular multi-agent system i, if the follower singular multi-agent system j is communicated with the follower singular multi-agent system i, then a ij =a ji >0; otherwise, a ij =0;b i For the connection weight between the follower singular multi-agent and the leader singular multi-agent, if the follower singular multi-agent system i is communicated with the leader singular multi-agent system, b i >0; otherwise, b i =0;
Step 3.2, according to the sliding mode control theory, when the proportion error system reaches the sliding mode surface,andestablishing; obtaining an equivalent controller u by taking the integral sliding mode surface derivative as zero eqi (t) is:
substituting the equivalent controller into a proportional error equation to obtain a sliding-mode kinetic equation:
step 3.3, designing the self-adaptive distributed controller based on the neural network
Wherein alpha is i Is the attack coefficient, σ, of the actuator under attack i (. Is) the input layer to hidden layer transfer function, ρ i (t)>0 is the neural network based adaptive distributed controller parameter to be solved,sgn (.) is a sign function, for an arbitrary function x, when x>At 0, sgn (x) =1; sgn (x) =0 when x =0; when x is<At 0, sgn (x) = -1;
the updating rule of the parameters is as follows:
wherein the matrixAndweight matrix gamma from hidden layer to output layer i And the weight Θ of the input layer to the hidden layer i Is determined by the estimation matrix of (a),is an unknown vector beta i Is determined by the estimated value of (c),is a Jacobian matrix, M i1 >0,M i2 >0,M i3 >0 is a gain matrix, vector
2. The method for proportional-tolerant tracking control of actuator attack singular multi-agent system as claimed in claim 1, wherein the model of the follower singular multi-agent system in step 1 is specifically:
consider N follower singular multi-agent systems containing nonlinear actuator attacks, where the model of the ith follower singular multi-agent system can be represented as
x i (t)=η i (t),t∈[-h,0] (1)
Wherein x i (t)∈R n Is the state of the ith agent, and h (t) is 0-h (t) -h<Infinity andtime-varying time lag of u ir (t)∈R p Is to control the transmissionIn, w i (t)∈R q Is an external disturbance; function eta i (t) is the initial value of agent i; the matrix pair (E, A) is regular; wherein A ∈ R n×n ,B u ∈R n×p And B w ∈R n×q Is a known system parameter matrix, matrix B u Column full rank;
the leader's singular multi-agent system model is:
x 0 (t)=η 0 (t),t∈[-h,0] (2)
wherein x 0 (t)∈R n Is the state of the leader's singular multi-agent system; initial state of leader singular multi-agent system is eta 0 (t);
The nonlinear actuator attack model containing time lag and the estimation method are as follows:
the actuator in the operation of the singular multi-agent system can be attacked, and the safety performance of the system is threatened: assuming that the attacking strength of actuators of each follower singular multi-agent system is different, defining an attack coefficient alpha i ∈[0,1](ii) a Then, the actuator attack function of the follower singular multi-agent system i can be expressed as:
u ir (t)=u i (t)+α i φ i (x i (t),x i (t-h(t)),t) (3)
wherein the content of the first and second substances,
using radial neural networksTo phi i (x i (t),x i (t-h (t)), t) estimation is performed with an estimation error ofWherein the vectorIs the input variable of the neural network with an input deviation of-1, vectorError vectorFor any e i >0 satisfiesMatrix ofRepresenting the weight of the hidden layer to the output layer,representing the weight from the input layer to the hidden layer; l i The number of nodes is hidden in the neural network, and p is the number of nodes in an output layer; nonlinear equation sigma i (. Is) an input-to-hidden-layer transfer function that can be expressed as a function vector of:wherein
For i =1,2.., N, the hypothetical matrixAndare respectivelyΓ i And Θ i The estimation matrix of (2); then, the estimation error of the weight matrix can be expressed asAndassuming a non-linear function phi i (x i (t),x i (t-h (t)), t) an estimation function of
Then, the error function:
is expressed as
Wherein, the first and the second end of the pipe are connected with each other,is a Jacobian matrix, and the infinitesimal term o (-) is expressed asNorm of residual termThe following conditions are satisfied:wherein beta is i ∈R 4 Is a vector that is not known and is,is beta i And the corresponding estimation error isFor an arbitrary matrix M ∈ R p×q ,Wherein λ is max (M T M) representation matrix M T The maximum eigenvalue of M;
3. the method for proportional-tolerant tracking control of an actuator attack singular multi-agent system as claimed in claim 2, wherein said step 4 is specifically:
wherein
The adaptive distributed controller (14) based on neural network can drive the state variables of the follower singular multi-agent system (1) to track the state of the leader singular multi-agent system (2) in a fixed proportion, i.e.,establishing all the singular multi-agent systems;
step 4.2 when w i When (t) ≠ 0, given gamma>0, if a matrix existsH 1 ∈R n×n Positive definite moment of symmetry Q 1 ∈R n×n ,H 1 ∈R n×n ,H 2 ∈R q×q Satisfy the requirements of
Wherein:
matrix U satisfiesIs a diagonal matrix and feeds back a gain matrixThe adaptive distributed controller (14) based on the neural network can drive the state variables of the follower singular multi-agent system (1) to track the state of the leader singular multi-agent system (2) according to a fixed proportion, and has a restraining effect on the external interference of the follower singular multi-agent system, namely,the external interference of the follower singular multi-agent system is satisfied when the external interference of all the singular multi-agent systems is satisfied:
4. the actuator attack singular multi-agent system scale tolerance tracking control method as claimed in claim 3, wherein said step 4.1 is embodied by:
step 4.1.1, adopting a singular system model decomposition method to carry out pair inequality A T X+X T A+Q 1 <0 is decomposed to obtain:
thus, det (A) 22 ) Not equal to 0,det (.) represents the determinant of the matrix, i.e. (E, A) no pulse;
step 4.1.2. Constructing the Lyapunov function of the formula (18):
wherein, the first and the second end of the pipe are connected with each other,
the Lyapunov function is derived along an error system under the action of a self-adaptive controller based on a neural network, and the following can be obtained:
matrix array l ij =-a ij All i is not equal to j; from the formula (16), it can be obtainedThus, error functionAndthe average index is stable; the proportional error tracking system has no pulse and satisfiesNamely, bagsA singular multi-agent system with time-lag nonlinear actuator attack realizes proportion-tolerant tracking control.
5. The method for proportional-tolerant tracking control of an actuator attack singular multi-agent system as claimed in claim 3, wherein said step 4.2 is embodied as:
when w (t) ≠ 0, combining the Lyapunov function constructed in the equation (18) with the following energy function:
solving the above equation can obtain:
wherein:
6. The method for proportional-tolerant tracking control of an actuator attack singular multi-agent system as claimed in claim 3, wherein said step 5 is specifically:
step 5.1. If the matrix K satisfies the condition in step 4.2 and the parameters in the adaptive controller (14) based on the neural network satisfy
Wherein, for all i =1, N,is a normal number; the integral sliding mode surface equation (11) can be reached in a limited time;
step 5.2, constructing a Lyapunov function of the formula (22):
is obtained by calculation
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211012274.3A CN115356929A (en) | 2022-08-23 | 2022-08-23 | Proportional allowable tracking control method for actuator attack singularity multi-agent system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211012274.3A CN115356929A (en) | 2022-08-23 | 2022-08-23 | Proportional allowable tracking control method for actuator attack singularity multi-agent system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115356929A true CN115356929A (en) | 2022-11-18 |
Family
ID=84003387
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211012274.3A Pending CN115356929A (en) | 2022-08-23 | 2022-08-23 | Proportional allowable tracking control method for actuator attack singularity multi-agent system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115356929A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116627042A (en) * | 2023-07-20 | 2023-08-22 | 南京邮电大学 | Distributed collaborative tracking method for asymmetric saturated multi-self-body system of actuator |
-
2022
- 2022-08-23 CN CN202211012274.3A patent/CN115356929A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116627042A (en) * | 2023-07-20 | 2023-08-22 | 南京邮电大学 | Distributed collaborative tracking method for asymmetric saturated multi-self-body system of actuator |
CN116627042B (en) * | 2023-07-20 | 2023-09-29 | 南京邮电大学 | Distributed collaborative tracking method for asymmetric saturated multi-self-body system of actuator |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109445447B (en) | Multi-agent formation tracking control method and system | |
Li et al. | Fuzzy adaptive optimal consensus fault-tolerant control for stochastic nonlinear multiagent systems | |
CN107422741B (en) | Learning-based cluster flight distributed attitude tracking control method for preserving preset performance | |
Hua et al. | Neural network observer-based networked control for a class of nonlinear systems | |
CN104950677A (en) | Mechanical arm system saturation compensation control method based on back-stepping sliding mode control | |
CN107193210B (en) | Adaptive learning preset performance control method of nonlinear system | |
Yang et al. | Formation control of multiple underwater vehicles subject to communication faults and uncertainties | |
CN112305918A (en) | Multi-agent system sliding mode fault-tolerant consistency control algorithm under supercoiled observer | |
CN106406097A (en) | Distributed adaptive coordinated control method for multi-manipulator systems | |
CN115356929A (en) | Proportional allowable tracking control method for actuator attack singularity multi-agent system | |
CN113885548B (en) | Multi-quad-rotor unmanned helicopter output constraint state consistent game controller | |
CN117055605A (en) | Multi-unmanned aerial vehicle attitude control method and system | |
CN111638648A (en) | Distributed pulse quasi-synchronization method with proportional delay complex dynamic network | |
Li et al. | Learning-observer-based adaptive tracking control of multiagent systems using compensation mechanism | |
CN114063457A (en) | Event-triggered fault-tolerant control method for mechanical arm system | |
Wei et al. | Adaptive finite‐time neural backstepping control for multi‐input and multi‐output state‐constrained nonlinear systems using tangent‐type nonlinear mapping | |
CN116755481A (en) | PID vibration active control and reliability analysis method based on neural network | |
CN110095989B (en) | Distributed multi-Lagrange system tracking control strategy based on back stepping method | |
Zhang et al. | Observer‐based consensus control for heterogeneous multi‐agent systems with intermittent communications | |
CN116000919A (en) | Full-state constraint control method of single-link mechanical arm system with dead zone | |
Niu et al. | Almost-global formation tracking control for multiple vehicles with disturbance rejection | |
CN113459083B (en) | Self-adaptive fixed time control method and system for mechanical arm under event trigger | |
Xia et al. | Distributed adaptive cooperative control via command filters for multi-agent systems including input unmodeled dynamics and sensor faults | |
Zhang et al. | Observer-Based Fuzzy Adaptive Formation Control for Saturated MIMO Nonlinear Multiagent Systems under Switched Topologies | |
CN111552317A (en) | Multi-spacecraft four-dimensional collaborative trajectory determination method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |