Fault Detection And Collaborative Control For Multi-Agent Systems Under Signed Networks

In recent years,the collaborative control of multi-agent systems has attracted the common attention of industry and academia due to its broad applications in many fields,such as intelligent transportation,multi-sensor cooperation,multi-robot formation.The essence of collaborative control of multi-agent systems is that multiple autonomous agents achieve an orderly collective behavior at the group level through sharing information and interacting with each other,and emerge a distributed and decentralized group intelligence.In addition,with the increasing scale and complexity of actual multi-agent systems,fault detection technology,as an important means to improve system safety and reliability,has become one of the issues that must be considered in the collaborative control process of the system.At present,most of the existing collaborative control theories and methods fo-cus on the multi-agent system with only cooperative interactions.However,in addition to cooperative interactions,competitive interactions are widespread in many practical multi-agent systems,such as economic systems,biological systems,social networks,robot sys-tems and so on.Such networks with both cooperative and competitive interactions are called signed networks.Compared with conventional cooperative networks,the emer-gence mechanism of group intelligence in multi-agent systems under signed networks is more complex,and the research on its fault detection mechanism is also more challeng-ing.In view of this,this dissertation takes the multi-agent system under signed networks as the research object,and conducts theoretical researches on its fault detection problem and several typical collaborative control problems.The main contents of this dissertation include:1.The problem of fault detection for multi-agent systems under signed networks with unknown input is studied,where the models with continuous-time dynamics and discrete-time dynamics are considered,respectively.For continuous-time multi-agent systems,a fault detection algorithm based on relative output information is proposed.The algo-rithm can be used to detect and isolate a single agent with actuator fault,and it realizes the distributed implementation of the algorithm.For discrete-time multi-agent systems,a fault detection algorithm based on H_-/H_∞unknown input observer is proposed.The algorithm not only ensures the sensitivity of the generated residual to the corresponding actuator fault,but also ensures the robustness to the undecouped unknown input.Finally,the effectiveness of the proposed fault detection algorithms is verified by numerical sim-ulations.2.The problem of bipartite consensus control for multi-agent systems under switch-ing signed networks is investigated.A distributed controller based on asynchronous com-munications is designed,where each agent only communicates with its neighbors at part of instants determined by the exclusive clock.In addition,the time-varying communica-tion delays are also considered in the design of the controller.The joint effects of asyn-chronous environments and communication delays pose challenges to the research of bi-partite consensus problem.To overcome challenges,by constructing mixed state vectors,the bipartite consensus problem of multi-agent systems is first equivalently transformed into the asymptotic stability problem of time-varying augmented system.Secondly,based on the elementary transformation of the matrix,an effective parameter selection strategy is proposed to ensure the nonnegativity of the coefficient matrix in the augmented sys-tem.Then,using the relevant properties of row-stochastic matrices,the convergence of the augmented system is analyzed,and sufficient criterion for the solvability of bipartite consensus problem is obtained.Finally,the theoretical result is verified by a numerical example.3.The problem of bipartite tracking control for multi-agent systems under signed networks with communication constraints is studied,where the cases subject to Do S at-tacks and asynchronous communications are considered,respectively.For the case of Do S attacks,a general Do S attack model is considered.Using the knowledge of nonneg-ative matrix theory and signed digraph theory,the convergence of multi-agent system is comprehensively analyzed,and the algebraic condition that depends on attack frequency and attack duration is obtained.For the case of asynchronous communications,a state feedback controller based on asynchronous communication mechanism is proposed.Us-ing the convergence analysis method of products of sub-stochastic matrices,the algebraic condition for realizing asynchronous bipartite tracking control of multi-agent system is presented.Numerical examples are provided to illustrate the correctness of the algebraic conditions.4.The problems of flocking control and bipartite flocking control for multi-agent sys-tems under signed networks are studied,where the cases with fixed topology and switch-ing topologies are considered,respectively.For flocking control,a class of signed net-works with general topology structure is considered,which allows competitive interac-tions between every pair of agents.By solving the convergence of the product of infi-nite super-stochastic matrices,the algebraic conditions for the solvability of the flocking control problem with fixed topology and switching topologies are presented.For bipar-tite flocking control,monotonically bounded positive and negative weight functions are constructed respectively to quantify the degree of cooperation and competition between agents.By solving the convergence of the product of infinite sub-stochastic matrices,the algebraic conditions for achieving bipartite flocking control under fixed topology and switching topologies are given.The correctness of the results obtained is demonstrated through several computer examples.5.The problem of bipartite containment control for multi-agent systems under switch-ing signed networks is studied,where the models with second-order dynamics and generic linear dynamics are considered,respectively.For second-order multi-agent systems,the system is equivalently transformed into a time-varying augmented system through appro-priate model transformation strategies.Using edge composition techniques,the conver-gence of time-varying augmented systems is analyzed,and the topological condition for achieving bipartite containment control of second-order multi-agent systems is given.For generic linear multi-agent systems,through the construction of an augmented digraph,the inequality constraint on the spectral radius of the system matrix is established,and the topological condition to ensure bipartite containment control of generic linear multi-agent systems is presented.The correctness of the theoretical results is verified through com-puter examples and experiments on UAV platform.