Research On Distributed Filtering And Control Of Multi-Agent Systems

During the last decade,distributed filtering and control of multi-agent systems has attracted considerable attention of researchers in the control and artificial intelligence communities.This is due to its wide applications in information fusion and target tracking of wireless sensor networks,formation control of multi-robot systems,congestion control of communication networks,analysis and optimization of socioeconomic systems,and so on.This dissertation focuses on distributed filtering and control of multi-agent systems and obtains the following research results:(1)The problem about strong tracking consensus filtering of multi-agent systems is studied.Based on the existing distributed filtering algorithms and the observation that strong tracking filter and current statistical model under model mismatches and inaccurate noise statistical characteristics can still track a target effectively,this dissertation proposes a distributed strong tracking consensus filtering algorithm based on the current statistical model,which can track a non-cooperative target effectively.The distributed estimation fusion algorithm can utilize the measurement information and redundancy of multiple sensors to improve the robustness and accuracy of parameter estimation,and provide a theoretical basis for multi-agent system to track a non-cooperative target.(2)The formation tracking problem on multi-quadrotor systems is investigated.Based on the model reference consensus protocol,the maximum distance consensus protocol and the reference controller,this dissertation proposes a formation control method with formation information feedback for multi-quadrotor systems.A distributed strong tracking consensus filter is used to modify the reference model consensus protocol and to guarantee that multi-quadrotor systems achieve the formation tracking for a non-cooperative target.(3)The problem on consensus tracking of second-order stochastic multi-agent systems based on sampled-data is considered.By introducing the acceleration approximation information of an agent and the relative delayed velocity information between an agent and a moving target,a sampled-data consensus tracking protocol for second-order multi-agent systems is proposed,which is suitable for the presence of measurement noises and a moving target with an asymptotically-unbounded acceleration and a bounded jerk.Using algebraic graph theory,eigenvalue analysis method and probability limit theory,the necessary and sufficient conditions ensuring the systems to achieve mean-square bounded consensus tracking are obtained.(4)The consensus problem of second-order discrete-time multi-agent systems based on event-driven is taken into account.In order to reduce the communication bandwidth of the network and save the energy consumption of sensors,a distributed event-driven consensus protocol is proposed.Based on Lyapunov function method,the sufficient conditions in the form of linear matrix inequalities are obtained to ensure consensus of second-order discrete-time multi-agent systems.The results show that the proposed event-driven consensus protocol is effective in ensuring consensus of second-order discrete-time multi-agent systems and reducing the network communication bandwidth.