Font Size: a A A

Research On Global Distributed Optimal Consensus Of Multi-Agent Systems

Posted on:2021-05-25Degree:DoctorType:Dissertation
Country:ChinaCandidate:H SunFull Text:PDF
GTID:1368330605972792Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Owing to their high efficiency,high fault tolerance and scalability,multi-agent systems have been widely used in many aspects,such as fighter escort,cooperative penetration of multiple missiles and intelligent traffic signal control.Consensus is a fundamental problem of multi-agent systems,which aims to design a control protocol such that the states of all agents converge to a same value.In practi-cal application,only achieving consensus can not meet the actual demand and cost function often needs to be taken into account to save resources.Especially when the scale of system is large and energy resources are limited,it is partic-ularly important to consider energy consumption.Therefore,optimal consensus of multi-agent systems has become one of the research directions under the spot-light in recent years.However,compared with single agent system,besides the dynamic equation of single agent,the complicated effects of the interaction be-tween agents on both the overall dynamics and the cost function for multi-agent systems are considered.Therefore,global optimization problem of multi-agent systems is more challenging.This dissertation investigates distributed optimal consensus for leader-following linear multi-agent systems.Particularly,global consensus cost function and com-munication topology are pre-given,and the optimal protocol is distributed(which merely depends on relative state information).This makes classical Riccati strat-egy in the existing literature inapplicable,mainly because that for a given global consensus cost function,merely a centralized optimal protocol can be derived by solving algebraic Riccati equation.In this dissertation,a novel framework of dis-tributed optimal consensus is developed by implementing the parameterization of cost function.Based on this,a distributed optimal protocol is designed for a pre-given global consensus cost function to achieve asymptotic consensus and minimization of cost function.The main works of this dissertation consist of the following three aspects:(?)Distributed optimal consensus of first-order multi-agent systemsGlobal distributed optimal consensus of typical multi-agent systems in single-integrator form is investigated.Based on the digraph of directed tree,a design strategy of distributed optimal consensus protocol is proposed by implementing the parameterization of cost function.Specifically,the explicit formula of the con-sensus error with respect to gain parameters is constructively presented.Then,an online-implementable recursive algorithm is established for the parameteriza-tion of cost function.Namely,the completely explicit formula of cost function depending on gain parameters is obtained.Based on this,the existence of opti-mal gain parameters is proven,which means the existence of distributed optimal protocol.Furthermore,the optimal gain parameters are derived by minimizing the explicit formula of cost function.(Chapter 3 in the dissertation)(?)Distributed optimal consensus of first-order multi-agent systems over more general topologyGlobal distributed optimal consensus of multi-agent systems in single-integrator form over a more general topology is investigated.In the topology under investi-gation,part of followers have two neighbors instead of each follower only having one neighbor,which makes the protocol design strategy over the digraph of direct-ed tree no longer applicable.To this end,the existence of the distributed optimal protocol(merely depending on relative state information)is first shown.Then,by developing a more general method of the parameterization of cost function,a novel strategy of distributed optimal consensus is established over the topology under investigation to achieve leader-following consensus and the minimization of global consensus cost function.(Chapter 4 in the dissertation)(III)Distributed optimal consensus of second-order multi-agent sys-temsGlobal distributed optimal consensus of second-order multi-agent systems is investigated.Essentially different from the case of first-order dynamics(with only achieving position consensus),both position consensus and velocity consensus are pursued in second-order systems.Correspondingly,the protocol should involve not only the relative position information but also the relative velocity informa-tion,which would give rise to inherent coupling between the dynamics of position and velocity of the closed-loop system for each agent.Moreover,the cost func-tional involved in the optimal consensus also contains both the relative position information and the relative velocity information,which would further increase the difficulty for achieving distributed optimal consensus by implementing the parameterization of cost function.For second-order multi-agent systems,the ex-istence of the minimum value of the cost function with respect to the gain param-eters over the digraph of directed tree is shown.Then,an online-implementable recursive algorithm is developed to accomplish the parameterization of consensus errors of relative position and relative velocity,and the parameterization of cost function.Furthermore,the optimal gain parameters are obtained by minimizing the explicit formula of cost function.(Chapter 5 in the dissertation)...
Keywords/Search Tags:Multi-agent systems, consensus, global optimization, distributed control, parameterization of cost function
PDF Full Text Request
Related items