Unmanned motorcade control method based on multi-agent fuzzy iterative learning control

Abstract

The invention relates to the field of unmanned vehicle control, and aims to solve the problems that a motorcade control method depends on a simplified unmanned control model, external environment interference is not fully considered, following vehicles cannot respond to sudden road conditions in real time, re-planning and re-layout are needed, and task interruption is caused; the invention provides an unmanned motorcade control method based on multi-agent fuzzy iterative learning control, and the method comprises the steps: obtaining the communication topology of a motorcade control system through the correlation between vehicles, determining a leader vehicle and a following vehicle of a motorcade according to a control target, dynamic characteristics, the communication topology and task demands, and carrying out the control of the leader vehicle and the following vehicle. Constructing an adjacent matrix and a Laplacian matrix; and a follower control law based on iterative learning is introduced into the motorcade control model to obtain a following control model of the unmanned motorcade, accurate following control of the unmanned motorcade in the presence of external disturbance is realized by using the following control model, and the state of the leader vehicle can be dynamically changed.

Description

Technical Field

[0001] The present invention relates to the field of unmanned vehicle control, and in particular to a method for controlling an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control. Background Art

[0002] With the advancement of science and technology, intelligence and automation are becoming increasingly prevalent. Driven by technologies such as the internet, artificial intelligence, and 5G, intelligent autonomous driving technology continues to develop. Vehicle platooning control is crucial for intelligent autonomous driving. Current research methods for autonomous platooning control primarily focus on distributed control, consensus algorithms, and artificial intelligence. After establishing a platoon control model, previously unrelated vehicles are connected to achieve platooning control within a platoon.

[0003] Through the above analysis, the problems and defects of the existing technology are as follows:

[0004] The current fleet control method is relatively simple, relying on a simplified unmanned vehicle fleet control model, and does not fully consider external environmental interference. During long-distance following, the error gradually increases with the number of iterations. For complex vehicle following control, the accuracy cannot meet the usage requirements.

[0005] Following vehicles can only follow the pre-set status of the leading vehicle. They cannot respond to sudden problems on the road in real time. The fleet needs to replan its layout, resulting in mission interruption.

[0006] Due to the inherent conflict between stability and optimality, the pursuit of stability will sacrifice energy consumption optimality and comfort, while the pursuit of optimality will choose low energy consumption, which is prone to cause oscillation. Existing methods cannot achieve the two goals of fleet stability and optimal control at the same time by integrating road data. Summary of the Invention

[0007] In response to the problems existing in the prior art, the present invention provides a method for controlling an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control. The present invention is suitable for controlling an unmanned vehicle fleet. By using a fully distributed fuzzy iterative learning controller with time-varying coupling gain, the operator indirectly controls the following vehicles by controlling the lead vehicle, thereby achieving precise following control of the fleet in the presence of external interference. Moreover, as the state of the lead vehicle changes in real time, the following vehicles can respond in real time, allowing the fleet to perform following control under complex road conditions.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for controlling an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control collects information about each vehicle in the fleet to be controlled and obtains the correlation between each vehicle, including the logical relationship and intensity of information interaction between vehicles. The communication topology of the unmanned vehicle fleet control system is obtained based on the correlation between vehicles. The leading vehicle and following vehicles of the fleet are determined according to the control objectives, dynamic characteristics, communication topology and task requirements of the unmanned vehicle fleet control system. The dynamic equations of the following vehicles and the leading vehicle are determined based on the correlation between each vehicle in the fleet, and an adjacency matrix and a Laplace matrix are constructed. The follower control law based on iterative learning is introduced into the fleet control model to obtain a following control model for the unmanned vehicle fleet, and the following control model is used to realize the control of the unmanned vehicle fleet.

[0010] Furthermore, the dynamic equations of the following vehicle and the input equations of the leading vehicle are obtained by the following operations:

[0011] The unmanned vehicle fleet consists of n follower vehicles and 1 leader vehicle, a mixed-order multi-agent system consisting of n+1 vehicles. Assuming that each unmanned vehicle has the characteristic of repeating operations in a finite time interval, the dynamic equation of vehicle i is:

[0012]

[0013] Where: i represents the i-th vehicle, x i (k,l)∈R 2 represents the state vector of the following vehicle, separates the platoon vehicles from α to distinguish the dynamic models of different orders for better cooperative control, and represents the different components of the state variables of the leading vehicle, which represent the pitch angle and pitch rate respectively, k∈{0,1,...,T} represents any time of repeated operation, l=1,2,..., represents the number of iterative learning, u i (k,l)∈R and ω i (k,l)∈R are the control input and external disturbance of the state vector of vehicle i at the kth moment and the lth iteration, respectively. i (x i (k,l)) is an unknown nonlinear function that satisfies the local Lipschitz condition;

[0014] The dynamic equation of the leading vehicle is:

[0015]

[0016] where x0∈R 2is the state vector of the leader vehicle, m0(x0,k) is the unknown nonlinear function of the leader vehicle, ω0(x0,k)∈R represents the bounded external disturbance that can be measured by the leader vehicle, and u0∈R is the unknown bounded control input, that is, there is an unknown positive constant and satisfy

[0017] Furthermore, the unknown nonlinear function m i (x i (k,l)) is described by the fuzzy logic system:

[0018]

[0019] in is the fuzzy weight vector, ν i (x i (k,l)) satisfies |ν i (x i (k,l))|≤M νi The fuzzy approximation error of M νi is a positive constant, Ω i and O i is a compact set; i (x i (k,l))=[φ i1 (x i (k,l)),...,φ il (x i (k,l))] T is the fuzzy basis function vector, where is defined as

[0020]

[0021] in and Φ ij are the center vector and width of the Gaussian function respectively.

[0022] Furthermore, the adjacency matrix and Laplace matrix are constructed as follows:

[0023] Nodes are used to represent different vehicles. The connection relationship between node i and node j is represented by the weighted adjacency matrix A. The diagonal elements a of the adjacency matrix A are i,j =0; if there is a connection between node i and node j, then a i,j =a j,i > 0, otherwise, a i,j =0;

[0024] The relationship between the vehicles is represented by the Laplace matrix L, and L = DA is defined;

[0025] The relationship between the leading vehicle and the following vehicles in the convoy is represented by a diagonal matrix E=diag{d1,...,d n} means, where d i Represents the relationship between vehicle i and the leader vehicle.

[0026] Furthermore, the states of the following vehicle and the leading vehicle satisfy the following conditions respectively:

[0027]

[0028] and

[0029] The sliding error based on the adjacent node information is defined as:

[0030]

[0031] Then we have: in

[0032] where r l =[r(1,l),...,r(n,l)] T ,

[0033] and Represent the Laplace matrices of the state communications corresponding to all vehicles.

[0034] Furthermore, the following control model of the unmanned vehicle fleet is constructed as follows:

[0035] In order to achieve the control goal, according to the dynamic equations of the leading vehicle and the following vehicle, the follower control law based on iterative learning is constructed as where u ia (k,l) represents the feedback controller, u ib (k, l) represents the use of adaptive fuzzy control to compensate for unknown nonlinear functions; u ic (k, l) represents the robust controller used to compensate for the approximation error as well as the unknown dynamics m0 of the leader vehicle and the unknown input u0, u id (k,l) represents the function used to compensate for external disturbances:

[0036] u ia (k,l)=g i (k,l)r i (k,l)

[0037]

[0038] where g i (k,l) is a time-varying adaptive gain, r i (k, l) is the sliding error based on the information of adjacent nodes; i is the fuzzy weight vector, φ i (x i (k,l)) is the fuzzy basis function vector, M i is a positive constant, and Respectively represent Ξ i 、M i and The estimated value, θ(k,l+1) is the same as that satisfying and Y>0, the constant parameter related to the number of iterations k, e i (k, l) represents the tracking error at time k in the lth iteration, K p and K d are proportional gain and differential gain respectively; is a time-varying vector that is bounded for any time k and number of iterations l.

[0039] Furthermore, the time-varying vector for:

[0040]

[0041] Among them, ζ>0, ξ>0 are variable parameters;

[0042] For any time k∈{0,1,...,T} and number of iterations l∈{0,1,...,T}, the perturbation ω i (k,l) are bounded, and each agent The sign is unchanged.

[0043] Furthermore, the sliding error is expressed as:

[0044]

[0045]

[0046] Here m(x l )=[m1(x 1,l ),...,m n,l (x n,l )] T ,u l =[u1,...,u n ] T , and 1 n =[1,...,1] T ∈R n ;

[0047] The communication topology consisting of the states of all the autonomous vehicles in the autonomous vehicle fleet is connected, i.e. and is a positive definite matrix.

[0048] In summary, the invention has the following beneficial effects:

[0049] Compared with the existing vehicle following technology, the present invention combines the iterative learning control method and fuzzy control theory to solve the problem of precise leadership and following consistency of mixed-order unknown nonlinear multi-agents with repeated operations, and realizes precise following control of unmanned vehicle fleets in the presence of external disturbances. In addition, the state of the leading vehicle can change dynamically. When the control protocol remains unchanged, the correlation of the vehicles in the unmanned vehicle fleet control system will not change, and the tracking error of the following vehicle is always bounded. For the control of the following vehicle, it is possible to improve the control accuracy and stability of the following vehicle while continuously adjusting the adjustable gain of the online learning parameters through trial and error method, so as to achieve the unity of real-time and optimality, which better meets the actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0051] Figure 1 It is a logical schematic diagram of the present invention.

[0052] Figure 2 It is a control flow chart of the present invention.

[0053] Figure 3 It is the communication topology set in the simulation experiment. Specific implementation methods

[0054] The drawings are for illustrative purposes only and are not to be construed as limiting the present invention.

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention, but are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without making creative work shall fall within the scope of protection of the present invention.

[0056] The present invention provides a method for controlling an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control. The method collects information on each vehicle in the fleet to be controlled, obtains the correlation between each vehicle, including the logical relationship and strength of information interaction between vehicles, and obtains the communication topology of the unmanned vehicle fleet control system based on the correlation between vehicles. The leading vehicle and following vehicles of the fleet are determined according to the control objectives, dynamic characteristics, communication topology and task requirements of the unmanned vehicle fleet control system; the dynamic equations of the following vehicles and the leading vehicle are determined according to the correlation between each vehicle in the fleet, and the adjacency matrix and the Laplace matrix are constructed; the follower control law based on iterative learning is introduced into the fleet control model to obtain a following control model of the unmanned vehicle fleet, and the following control model is used to realize the control of the unmanned vehicle fleet.

[0057] The dynamic equations of the following vehicle and the input equations of the leading vehicle are obtained as follows:

[0058] The unmanned vehicle fleet consists of n follower vehicles and 1 leader vehicle, a mixed-order multi-agent system consisting of n+1 vehicles. Assuming that each unmanned vehicle has the characteristic of repeating operations in a finite time interval, the dynamic equation of vehicle i is:

[0059]

[0060] Where: i represents the i-th vehicle, x i (k,l)∈R 2 represents the state vector of the following vehicle, separates the platoon vehicles from α to distinguish the dynamic models of different orders for better cooperative control, and represents the different components of the state variables of the leading vehicle, which represent the pitch angle and pitch rate respectively, k∈{0,1,...,T} represents any time of repeated operation, l=1,2,..., represents the number of iterative learning, u i (k,l)∈R and ω i (k,l)∈R are the control input and external disturbance of the state vector of vehicle i at the kth moment and the lth iteration, respectively. i (x i (k,l)) is an unknown nonlinear function that satisfies the local Lipschitz condition;

[0061] The dynamic equation of the leading vehicle is:

[0062]

[0063] where x0∈R 2is the state vector of the leader vehicle, m0(x0,k) is the unknown nonlinear function of the leader vehicle, ω0(x0,k)∈R represents the bounded external disturbance that can be measured by the leader vehicle, and u0∈R is the unknown bounded control input, that is, there is an unknown positive constant and satisfy

[0064] Unknown nonlinear function m i (x i (k,l)) is described by the fuzzy logic system:

[0065]

[0066] in is the fuzzy weight vector, ν i (x i (k,l)) satisfies |ν i (x i (k,l))|≤M νi The fuzzy approximation error of M νi is a positive constant, Ω i and O i is a compact set; i (x i (k,l))=[φ i1 (x i (k,l)),...,φ il (x i (k,l))] T is the fuzzy basis function vector, where is defined as

[0067]

[0068] in and Φ ij are the center vector and width of the Gaussian function respectively.

[0069] The adjacency matrix and Laplace matrix are constructed as follows:

[0070] Nodes are used to represent different vehicles. The connection relationship between node i and node j is represented by the weighted adjacency matrix A. The diagonal elements a of the adjacency matrix A are i,j =0; if there is a connection between node i and node j, then a i,j =a j,i > 0, otherwise, a i,j =0;

[0071] The relationship between the vehicles is represented by the Laplace matrix L, and L = DA is defined;

[0072] The relationship between the leading vehicle and the following vehicles in the convoy is represented by a diagonal matrix E=diag{d1,...,d n} means, where d i Represents the relationship between vehicle i and the leader vehicle.

[0073] The states of the following vehicle and the leading vehicle satisfy the following conditions respectively:

[0074]

[0075] and

[0076] The sliding error based on the adjacent node information is defined as:

[0077]

[0078] Then we have: in

[0079] in

[0080]

[0081] and Represent the Laplace matrices of the state communications corresponding to all vehicles.

[0082] The following control model construction method of the unmanned vehicle fleet is as follows:

[0083] In order to achieve the control goal, according to the dynamic equations of the leading vehicle and the following vehicle, the follower control law based on iterative learning is constructed as where u ia (k,l) represents the feedback controller, u ib (k, l) represents the use of adaptive fuzzy control to compensate for unknown nonlinear functions; u ic (k, l) represents the robust controller used to compensate for the approximation error as well as the unknown dynamics m0 of the leader vehicle and the unknown input u0, u id (k,l) represents the function used to compensate for external disturbances:

[0084] u ia (k,l)=g i (k,l)r i (k,l)

[0085]

[0086] where g i(k,l) is a time-varying adaptive gain, r i (k, l) is the sliding error based on the information of adjacent nodes; i is the fuzzy weight vector, φ i (x i (k,l)) is the fuzzy basis function vector, M i is a positive constant, and Respectively represent Ξ i 、M i and The estimated value, θ(k,l+1) is the same as that satisfying and Y>0, the constant parameter related to the number of iterations k, e i (k, l) represents the tracking error at time k in the lth iteration, K p and K d are proportional gain and differential gain respectively; is a time-varying vector that is bounded for any time k and number of iterations l.

[0087] Time-varying vector for:

[0088]

[0089] Among them, ζ>0, ξ>0 are variable parameters;

[0090] For any time k∈{0,1,...,T} and number of iterations l∈{0,1,...,T}, the perturbation ω i (k,l) are bounded, and each agent The sign is unchanged.

[0091] The sliding error is expressed as:

[0092]

[0093] Here m(x l )=[m1(x 1,l ),...,m n,l (x n,l )] T ,u l =[u1,...,u n ] T , and 1 n =[1,...,1] T ∈R n ;

[0094] The communication topology consisting of the states of all the autonomous vehicles in the autonomous vehicle fleet is connected, i.e. and It is a positive definite matrix. The key state information directly related to the control target, including dynamic state, input state, neighbor interaction state, adaptive parameter state, sliding film error and other state information, can be used as state information.

[0095] like Figure 1 As shown in the figure, the method specifically includes four steps:

[0096] The first step is to identify a leader vehicle and at least one follower vehicle in the driverless car convoy.

[0097] An unmanned vehicle convoy consists of at least two unmanned vehicles. The leading vehicle's driving control is independent of the following vehicles, and the following vehicles will follow the leading vehicle. Therefore, the leading vehicle's driving speed and direction determine the driving speed and direction of the unmanned vehicle convoy.

[0098] It should be noted that the leading vehicle determines the driving status of the unmanned vehicle fleet, and the control model of the leading vehicle is constructed according to the actual needs of the fleet.

[0099] The second step is to determine the dynamic equations of the following vehicle and the leading vehicle.

[0100] The unmanned vehicle fleet consists of n follower vehicles and 1 leader vehicle, a mixed-order multi-agent system consisting of n+1 vehicles, namely mixed-order MASs. It is assumed that each unmanned vehicle has the characteristic of repeated operation in a finite time interval.

[0101] Construct the dynamic equations of the i-th unmanned vehicle and the dynamic equations of the leader vehicle.

[0102] The third step is to determine the connection relationship between vehicles, that is, to determine the dynamic equations of the following vehicle and the leading vehicle based on the correlation between the vehicles, and to construct the adjacency matrix and Laplace matrix.

[0103] Nodes are used to represent different vehicles. The connection between node i and node j is represented by the weighted adjacency matrix A. The diagonal element a of the weighted adjacency matrix A is i,j = 0, if there is a connection between node i and node j, then a i,j =a j,i > 0, otherwise, a i,j =0.

[0104] The relationship between vehicles is represented by the Laplace matrix L, and L=DA is defined.

[0105] The relationship between the leading vehicle and the following vehicles in the convoy is represented by a diagonal matrix E=diag{d1,...,d n} means, where d i represents the relationship between the following vehicle i and the leading vehicle.

[0106] In the fourth step, to achieve the control objective, a follower control law based on iterative learning is constructed based on the dynamic equations of the leading and following vehicles. This follower control law based on iterative learning is introduced into the convoy control model to obtain the following control model for the unmanned convoy. The following control model is used to realize the control of the unmanned convoy, where:

[0107] The follower control law is in

[0108] u ia (k,l)=g i (k,l)r i (k,l)

[0109]

[0110] where g i (k,l) is a time-varying adaptive gain, r i (k,l) is the sliding error based on the information of adjacent nodes.

[0111] Ξ i is the fuzzy weight vector, φ i (x i (k,l)) is the fuzzy basis function vector.

[0112] M i is a positive constant, and Respectively represent Ξ i 、M i and The estimated value, θ(k,l+1) is the same as that satisfying A constant parameter related to the number of iterations k of Y(>0), e i (k, l) represents the tracking error at time k in the lth iteration, K p and K d are the proportional gain and the differential gain respectively.

[0113] is a time-varying vector that is bounded for any time k and number of iterations l.

[0114] The convoy following model is as follows Figure 2 shown.

[0115] First, a communication topology matrix is ​​constructed from the information of the following and leading vehicles. Each node represents a different vehicle, and the matrix shows the relationship between the vehicles. The errors between the nodes are derived from the topology matrix and substituted into the parameter update rule and iterative learning controller of the following control model to control the following vehicle. This error is then substituted into the parameter update rule and iterative learning controller of the following control model in a loop to achieve following control.

[0116] like Figure 3 This can be viewed as the communication topology of a driverless car fleet, where 0 represents the leader vehicle, 1, 2, 3, and 4 represent the following vehicles, and only 1 is directly associated with the leader vehicle.

[0117] The dynamic equation of the follower in the system is described as

[0118]

[0119] Where: i represents the i-th vehicle, x i (k,l)∈R 2 Represents the state vector of the following vehicle. Some vehicles need to precisely control their position and speed (second order) in actual use, while other vehicles only need to adjust their speed (first order). Through this separation, the present invention can more efficiently achieve the cooperative control goal by separating the platoon vehicles from α to distinguish the dynamic models of different orders to better achieve cooperative control. and represents the different components of the state variables of the leading vehicle, which represent the pitch angle and pitch rate respectively, k∈{0,1,...,T} represents any time of repeated operation, l=1,2,..., represents the number of iterative learning, u i (k,l)∈R and ω i (k,l)∈R are the control input and external disturbance of the state vector of vehicle i at the kth moment and the lth iteration, respectively. i (x i (k,l)) is an unknown nonlinear function that satisfies the local Lipschitz condition.

[0120] The dynamic equations of the leading vehicle are described as:

[0121]

[0122] where x0∈R 2 is the state vector of the leader vehicle, m0(x0,k) is the unknown nonlinear function of the leader vehicle, ω0(x0,k)∈R represents the bounded external disturbance that can be measured by the leader vehicle, and u0∈R is the unknown bounded control input, that is, there is an unknown positive constant and satisfy

[0123] Set the initial state of the follower:

[0124] For the following vehicle i, the follower control law based on iterative learning is constructed as where u ia (k,l) represents the feedback controller, u ib (k, l) represents the use of adaptive fuzzy control to compensate for unknown nonlinear functions; u ic (k,l) represents the robust controller used to compensate for the approximation error, as well as the unknown dynamics m0 of the leader and the unknown inputs u0, u id (k,l) represents the function used to compensate for external disturbances:

[0125] u ia (k,l)=g i (k,l)r i (k,l)

[0126]

[0127] Set the parameters of the PD controller: proportional control item K p By adjusting the error between the current state of the leading vehicle and the following vehicle, the following vehicle is forced to move closer to the state of the leading vehicle; the differential control term K d According to the error of the state change rate, the changing trend of the following vehicles is adjusted to enhance the stability of the unmanned vehicle fleet control system.

[0128] Use the fuzzy logic system to approximate the unknown nonlinear function of the following vehicle i and construct the fuzzy logic system parameters

[0129] in and Φ ij They are the center vector and width of the Gaussian function, respectively. In the following control of the vehicle, according to the control target and road conditions, the and Φ ij These two parameters enable the unmanned vehicle fleet control system to strike a balance between stability and response speed.

[0130] Set the adjustable gain of online learning parameters: δ, θ l ,σ.

[0131] δ determines the estimated values ​​of the fuzzy logic system parameters The update speed of the parameter increases. When δ increases, the parameter update step size becomes larger, allowing the autonomous convoy control system to more quickly adjust parameters based on the state error between the current following and leading vehicles. When the vehicle driving environment suddenly changes, causing the following error to increase, a larger δ allows the controller to adjust quickly, accelerating the follower vehicle's convergence to the leader vehicle's state and improving the responsiveness of the convoy control system. However, an excessively large δ may amplify noise. The specific range of δ needs to be determined based on the control objectives, dynamic characteristics, and stability requirements of the autonomous convoy control system.

[0132] σ constrains the parameter estimates by introducing a negative feedback term related to the parameter estimates. The growth rate of Maintain a balance between the resolution and stability of the unmanned vehicle fleet system, and ensure that the control objectives are both within a reasonable range of resolution and stability.

[0133] where θ l is satisfied Parameters related to the number of iterations l, Y(>0) is a constant; δ is a constant.

[0134] This invention combines iterative learning control (ILC) methods with fuzzy control theory for controlling unmanned vehicle fleets, enabling formation control and enabling human operators to participate in the coordinated control of the fleet. Furthermore, all following vehicles are unaware of the lead vehicle's inputs. By utilizing a fully distributed fuzzy iterative learning controller with time-varying coupling gain, precise following control of the fleet is achieved, even in the presence of external interference. Furthermore, the operator directly controls the state of the lead vehicle and indirectly controls the following vehicles, resulting in a control process that exhibits both real-time and optimal performance.

[0135] While the embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to these specific embodiments. The above embodiments are merely illustrative and are not intended to be limiting. Those skilled in the art, informed by the present invention, may make various changes and adjustments within the spirit and scope of the claims, and such changes and adjustments shall be considered part of the scope of protection of the present invention.

Claims

1. A control method for an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control, characterized by: Collect information about each vehicle in the convoy to be controlled and obtain the correlation between each vehicle, including the logical relationship and strength of information exchange between vehicles. Use the correlation between vehicles to obtain the communication topology of the unmanned convoy control system. Determine the leading and following vehicles in the convoy based on the control objectives, dynamic characteristics, communication topology, and mission requirements of the unmanned convoy control system. According to the correlation between each vehicle in the convoy, the dynamic equations of the following vehicle and the leading vehicle are determined, and the adjacency matrix and the Laplace matrix are constructed; and the follower control law based on iterative learning is introduced into the convoy control model to obtain the following control model of the unmanned vehicle convoy, and the following control model is used to realize the control of the unmanned vehicle convoy.

2. The method for controlling an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control according to claim 1, characterized in that: The dynamic equations of the following vehicle and the input equations of the leading vehicle are obtained by the following operations: The unmanned vehicle fleet consists of n follower vehicles and 1 leader vehicle, a mixed-order multi-agent system consisting of n+1 vehicles. Assuming that each unmanned vehicle has the characteristic of repeating operations in a finite time interval, the dynamic equation of vehicle i is: Where: i represents the i-th vehicle, x i (k,l)∈R 2 represents the state vector of the following vehicle, separates the platoon vehicles from α to distinguish the dynamic models of different orders for better cooperative control, and represents the different components of the state variables of the leading vehicle, which represent the pitch angle and pitch rate respectively, k∈{0,1,...,T} represents any time of repeated operation, l=1,2,..., represents the number of iterative learning, u i (k,l)∈R and ω i (k,l)∈R are the control input and external disturbance of the state vector of vehicle i at the kth moment and the lth iteration, respectively. i (x i (k,l)) is an unknown nonlinear function that satisfies the local Lipschitz condition; The dynamic equation of the leading vehicle is: where x0∈R 2 is the state vector of the leader vehicle, m0(x0,k) is the unknown nonlinear function of the leader vehicle, ω0(x0,k)∈R represents the bounded external disturbance that can be measured by the leader vehicle, and u0∈R is the unknown bounded control input, that is, there is an unknown positive constant and satisfy 3. The human-involved fleet control method based on fuzzy iterative learning control according to claim 2, characterized in that: The unknown nonlinear function m i (x i (k,l)) is described by the fuzzy logic system: in is the fuzzy weight vector, ν i (x i (k,l)) satisfies |ν i (x i (k,l))|≤M νi The fuzzy approximation error of M νi is a positive constant, Ω i and O i is a compact set; i (x i (k,l))=[φ i1 (x i (k,l)),...,φ il (x i (k,l))] T is the fuzzy basis function vector, where is defined as in and Φ ij are the center vector and width of the Gaussian function respectively.

4. The human-involved fleet control method based on fuzzy iterative learning control according to claim 1, characterized in that: The adjacency matrix and Laplace matrix construction method are as follows: Nodes are used to represent different vehicles. The connection relationship between node i and node j is represented by the weighted adjacency matrix A. The diagonal elements a of the adjacency matrix A are i,j =0; If there is a connection between node i and node j, then a i,j =a j,i > 0, otherwise, a i,j =0; The relationship between the vehicles is represented by the Laplace matrix L, and L = DA is defined; The relationship between the leading vehicle and the following vehicles in the convoy is represented by a diagonal matrix E=diag{d1,...,d n } means, where d i Represents the relationship between vehicle i and the leader vehicle.

5. The human-involved fleet control method based on fuzzy iterative learning control according to claim 2 or claim 4, characterized in that: The states of the following vehicle and the leading vehicle respectively satisfy the following conditions: and The sliding error based on the adjacent node information is defined as: Then we have: in where r l =[r(1,l),...,r(n,l)] T , and Represent the Laplace matrices of the state communications corresponding to all vehicles.

6. The method for controlling an unmanned vehicle fleet based on multi-agent fuzzy iterative learning control according to claim 1, characterized in that: The following control model construction method of the unmanned vehicle fleet is as follows: In order to achieve the control goal, according to the dynamic equations of the leading vehicle and the following vehicle, the follower control law based on iterative learning is constructed as where u ia (k,l) represents the feedback controller, u ib (k, l) represents the use of adaptive fuzzy control to compensate for unknown nonlinear functions; u ic (k, l) represents the robust controller used to compensate for the approximation error as well as the unknown dynamics m0 of the leader vehicle and the unknown input u0, u id (k,l) represents the function used to compensate for external disturbances: u ia (k,l)=g i (k,l)r i (k,l) where g i (k,l) is a time-varying adaptive gain, r i (k, l) is the sliding error based on the information of adjacent nodes; i is the fuzzy weight vector, φ i (x i (k,l)) is the fuzzy basis function vector, M i is a positive constant, and Respectively represent Ξ i 、M i and The estimated value, θ(k,l+1) is the same as that satisfying and Y>0, the constant parameter related to the number of iterations k, e i (k, l) represents the tracking error at time k in the lth iteration, K p and K d are proportional gain and differential gain respectively; is a time-varying vector that is bounded for any time k and number of iterations l.

7. The human-involved fleet control method based on fuzzy iterative learning control according to claim 6, characterized in that: The time-varying vector for: Among them, ζ>0, ξ>0 are variable parameters; For any time k∈{0,1,...,T} and number of iterations l∈{0,1,...,T}, the perturbation ω i (k,l) are bounded, and each agent The sign is unchanged.

8. The human-involved fleet control method based on fuzzy iterative learning control according to claim 6, characterized in that: The sliding error is expressed as: Here m(x l )=[m1(x 1,l ),...,m n,l (x n,l )] T ,u l =[u1,...,u n ] T , and 1 n =[1,...,1] T ∈R n ; The communication topology consisting of the states of all the autonomous vehicles in the autonomous vehicle fleet is connected, i.e. and is a positive definite matrix.

Images