Research On Cooperative Control Of Networked Multi-agent Systems Under Stochastic Environments

Multi-agent systems have been widely used in real life and have become hot issue of research.It is a system which is composed of multiple agents that can communicate with each other and interact through cooperation and information exchange.The communication network is introduced into multi agent systems,which promotes the development of cooperative control of networked multi-agent system.In practice,however,multi-agent systems will be affected by some random factors,such as random noise,the random disturbance,the problem of limited transmission speed,time delay of information exchange caused by some hardware problems,etc.Thus it is of great theoretical and practical significance to study the cooperative control problems of networked multi-agent systems under the random environment.In this paper,several kinds of cooperative control problems of networked multi-agent systems in random environment are studied.The main contents are as follows:●The containment control problem of stochastic multi-agent systems with semi-Markovian switching topology is studied.In a more specific case,the distribution function of dwell time depends on both the current topology mode and the target mode.Considering the random noise and semiMarkovian switching topology,the sufficient conditions for achieving the control objective in the mean square sense are obtained.●The H∞ containment control problem of stochastic nonlinear multi-agent systems under semi-Markovian switching topology is investigated.By introducing the distribution function and probability density function of dwell time,a lemma of the specific formula of transition rate function is given,and they depend on the current mode and the next mode of topology.This formula is applicable to the relevant calculations of more general models.In terms of topological conditions,we relax the topological assumption that each switching topology needs to contain a spanning tree in the relevant references,and only requires the union graph of all possible topologies contains a spanning tree.Finally,the sufficient conditions for achieving H∞containment control in the mean square sense are given.●The leader-following consensus of general linear multi-agent systems with random switching topology is studied.Dwell time on each topology mode includes two parts:fixed part and random part.In random part,the dwell time obeys semi-Markovian process.Namely,when the system switches to a topology mode,it will first stay in this mode for a fixed time,and then switch to the next topology mode according to the semi-Markovian process.This kind of switch pattern can avoid the frequent switching and therefore is more general and practical.Finally,the sufficient and necessary conditions for achieving the control target are obtained.The sufficient and necessary conditions for achieving leader-following consensus under semi-Markovian switching topology can also be obtained without considering the fixed part dwelling in each topology.●The H∞ leader-following consensus of multi-agent systems with random noise and random input delay is studied by designing a new predictor-based controller.The input delay satisfies Bernoulli distribution,and the input delay can change fast over time,and the rate of change can be greater than 1.By using Lyapunov function method and stochastic knowledge,H∞leader-follow consensus in the mean square sense is finally achieved and the sufficient conditions are obtained.●The time-varying formation control of general linear multi-agent systems with some leaders in a directed topology is studied,and a hybrid dynamic event-triggered mechanism is proposed.In this mechanism,the parameter h>0 is introduced.That is,after sending the measurement data,the sensor will wait for at least time h,and then begins to check whether the triggered conditions is satisfied and decides wether to continue transmitting the measurement data.It can be seen that h is the lower bound of the time interval between two triggering,so no Zeno behavior will occur for each agent.Considering that too large h is not conducive to the realization of the control goal,thus the range of the parameter h is discussed.In order to save resources,an algorithm based on intermittent communication instead of continuous communication is designed.The improved dynamic event-triggered mechanism is more general and superior than the dynamic event-triggered mechanism.In the same time,the triggering times of hybrid dynamic event-triggered mechanism can be effectively reduced,which can reduce the consumption of equipment and save resources.