CN-121978915-A - Unmanned aerial vehicle cluster system distributed prediction control method under hybrid network attack

CN121978915ACN 121978915 ACN121978915 ACN 121978915ACN-121978915-A

Abstract

The invention discloses a distributed predictive control method for an unmanned aerial vehicle cluster system under hybrid network attack. Firstly, decoupling a global cost function, and approximately converting the distributed predictive control problem of the unmanned aerial vehicle cluster into a centralized predictive control architecture in which each unmanned aerial vehicle independently operates. On the basis, a state observer is designed offline to estimate the state of the system, and a multi-step control invariant set is introduced based on the estimated state to construct a distributed predictive control strategy. Further, a health judging mechanism is designed, the neighbor relation among unmanned aerial vehicles is dynamically updated, and nodes suffering from network attack are isolated from the communication topology in a fault isolation mode. Meanwhile, by introducing compatibility constraint, the asymptotic stability of the global system is ensured. Finally, the feasibility and the superiority of the method under the environment of the hybrid network attack are verified by combining simulation and physical experiments.

Inventors

TANG XIAOMING
LIU YANGJIE
CAI LINQIN
WANG HUIMING
LI HAOLAN
CAO FUSHUN
QIN QING

Assignees

重庆邮电大学

Dates

Publication Date: 20260505
Application Date: 20260106

Claims (7)

1. The distributed predictive control method for the unmanned aerial vehicle cluster system under the hybrid network attack is characterized by comprising the following steps of: Step a, offline design of a state observer is carried out, and the state of a system is estimated; Step b, introducing a multi-step control invariant set design distributed control strategy; C, constructing a compatibility constraint condition to enable the global system to be asymptotically stable; and d, updating the neighbor relation among unmanned aerial vehicles through a health judging mechanism, and isolating the nodes suffering from network attack from formation communication by adopting fault isolation.
2. The method for distributed predictive control of a cluster system of a hybrid network attack according to claim 1, wherein in the step a, a mathematical model of the cluster system of the unmanned plane is established as follows: consider a drone cluster system that can be decoupled into M subsystems, where the state space equation for drone i is: Representing the state and output of the system and the control input, respectively; is a known constant matrix; the hybrid attack occurring in the communication network between the controller and the actuator is described as: representing the control input before attack, 、 Representing the probability of attack, Represents an intermediate variable; The specific expression of the observer and the controller is as follows: Wherein, the , The observed value of the system state and the output of the observer, respectively; And Respectively representing observer gain and feedback gain; A value of x at k+t representing a prediction at k, The representation set {0,1, N-1}, N being a specified scalar; design of reference centralized predictive control objective Divided into Wherein For UAVs The control targets are as follows: ; representing the set {0,1, M } Respectively represent the state matrix of the adjacent unmanned aerial vehicle and the current unmanned aerial vehicle, Representing a weight matrix, Is an intermediate variable.
3. The method for distributed predictive control of a cluster system of unmanned aerial vehicles under hybrid network attack of claim 2, wherein the method comprises obtaining an observer gain On the basis of the above, designing a distributed model predictive controller based on a multi-step control invariant set; the design based on the multi-step control invariant set is divided into the following steps: i. Introducing the system state into an ellipsoid invariant set: is a group of ellipsoids invariable set, 、、 Is an intermediate matrix variable; defining a time-varying Lyapunov function and forcing the Lyapunov function to decrement and satisfy Representing the lyapunov function, 、 Representing a weight matrix; define a time-varying lyapunov function: Wherein, the , , , 、、 Is an intermediate matrix variable; , Is a specified scalar quantity which is set to the value of the target, The observation error is that: Thus, the first and second substrates are bonded together, Can use a finite time domain cost function To approximate; by multi-step control set Introducing system state into control invariant set The conditions of (2) are: is an intermediate matrix variable; defining an upper limit of the maximum value of the objective function as follows: Wherein the method comprises the steps of To finally control the objective cost function, Is an intermediate matrix variable, Is a weight matrix.
4. The method for controlling the distributed prediction of the unmanned aerial vehicle cluster system under the hybrid network attack of claim 3, wherein sufficient conditions for ensuring asymptotic stability of the global system are considered for designing compatibility constraints; Whenever a neighbor UAV At the time of When solving the optimization problem, let When (when) Can obtain . Is in a feasible state, Is in the optimal state of the system, For a specified scalar, Is the optimal system control input.
5. The method for distributed predictive control of a cluster system of unmanned aerial vehicles under hybrid network attack according to claim 4, wherein the step d updates the communication relationship in the unmanned aerial vehicle cluster by using a health discrimination mechanism, specifically comprises: firstly abstracting unmanned aerial vehicle cluster formation into a plurality of nodes, and then setting the weight of the edge connected with the communication fault node to be zero by utilizing a health discrimination mechanism; Firstly, updating the communication relation between unmanned aerial vehicles through a health judging mechanism, isolating nodes suffering from network attack from formation communication by adopting fault isolation, and describing the nodes as Wherein, the , The communication before encountering an attack is described, The communication after encountering an attack is described, Indicating that the communication relationship was prior to encountering the attack, The value of (2) is determined by a health discrimination mechanism; the health discrimination mechanism can be described as follows: Step1 at At moment, unmanned aerial vehicle Detecting the health state of the host machine, if the host machine is in a complete health state, the host machine can be arranged Otherwise ; Step2, when the unmanned aerial vehicles transmit information, namely the position and speed information of each unmanned aerial vehicle according to communication, the health information can be obtained Adding to the transferred information; step3 unmanned plane Receiving unmanned plane If the information of (1) Put in If (1) I.e. comprising Or (b) I.e. no information is received, then the device is set ; Step4 output Value of (1) order Returning to Step1; and finally, designing a cluster control algorithm under network attack based on a distributed predictive control theory according to a communication fault processing strategy and a control scheme.
6. An electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for distributed predictive control of a cluster system of unmanned aerial vehicles under a hybrid network attack as claimed in any one of claims 1 to 5 when the program is executed by the processor.
7. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the method for distributed predictive control of a drone cluster system under hybrid network attack according to any of claims 1 to 5.

Description

Unmanned aerial vehicle cluster system distributed prediction control method under hybrid network attack Technical Field The invention relates to the field of unmanned aerial vehicle cluster system control, in particular to a method for predicting and controlling an unmanned aerial vehicle cluster system under hybrid network attack. Background The unmanned aerial vehicle cluster system is an intelligent task system formed by autonomous networking, collaborative decision-making and distributed control of a plurality of unmanned aerial vehicles, and has the core that global task coordination is realized by means of local information interaction, and the intelligent task system has the remarkable advantages of strong environmental adaptability, high system robustness, good task expandability and the like. As a new generation of strategic emerging field of artificial intelligence and aerospace technology integration, unmanned aerial vehicle clusters are not only the key capability support of national defense safety, but also new kinetic energy for promoting economic and social development, and have important application potential in various scenes such as intelligent logistics, precise agriculture, smart cities and the like. At present, unmanned aerial vehicle cluster system research is still in a continuous development stage, and the safety and reliability of the unmanned aerial vehicle cluster system research are main bottlenecks for restricting the large-scale application of the technology. Particularly, various network attacks and network constraints existing in a communication network form a serious challenge for the stable operation of the unmanned aerial vehicle cluster system. The existing research does not fully cover the problem of unmanned aerial vehicle cluster distributed predictive control under the hybrid network attack, and a related systematic method is still lacking. In addition, the multi-step control invariant set method is not effectively applied to unmanned aerial vehicle cluster control, and the method can improve the control performance of the system through multi-step feedback gain, so that a new solution idea is provided for the problems. Aiming at the blank of the research, the research provides a hybrid network attack-oriented unmanned aerial vehicle cluster distributed predictive control method to enhance the stability and reliability of the system under complex interference. Through retrieval, application publication number CN118192661A, a distributed collaborative prediction control method of unmanned aerial vehicles facing a communication damage scene is provided, a multi-unmanned aerial vehicle system model under collaborative flight tasks is established, a network model under the communication damage scene and corresponding transmission channel buffer models are configured for each unmanned aerial vehicle, an improved robust multivariate observer is used for estimating the disturbed state of the unmanned aerial vehicle and the interference information estimation in an output channel, the estimated state is used as the input of a distributed collaborative prediction controller and compensating corresponding interference, a control input generation and compensation strategy facing the communication damage scene is obtained, and the distributed collaborative prediction control of the unmanned aerial vehicles facing the communication damage scene is realized. The invention effectively filters the measurement disturbance and FDI attack influence in the output channel and obtains the estimated value of the actual state, effectively has the communication damage phenomenon of random packet loss of the transmission channel, reduces the calculation burden in the collaborative flight task, and has higher reliability under the unknown interference and complex network environment. CN118192661A adopts a coupled distributed architecture, a global cost function is not decoupled, nodes need to be solved in a combined optimization mode, so that the computational complexity is high, iteration is easy to be not converged when communication is damaged, and the real-time requirement of a large-scale cluster is difficult to adapt. According to the invention, the cluster control problem is converted into independent centralized predictive control of each node through decoupling the global cost function, and the local information is only relied on for solving, so that the calculation and communication burden is greatly reduced, the control instantaneity in a communication impaired scene is ensured, and the large-scale deployment applicability is improved. In the aspect of network attack handling, the CN118192661A only passively filters interference and FDI attack influence through an observer, and the lack of an attack node identification and isolation mechanism easily causes error information diffusion and global instability. The invention designs a health judging mechanism, accurately identifies the