Font Size: a A A

On Event-Triggered Control And Distributed Optimization For Consensus Of Multi-Agent Systems

Posted on:2020-11-16Degree:DoctorType:Dissertation
Country:ChinaCandidate:L JianFull Text:PDF
GTID:1368330623958198Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
The consensus behavior of multi-agent systems is a hot research topic in the sys-tems and control community in recent years.It can be applied in many practical fields,such as multi-aircraft system formation control,smart grid resource allocation,and multi-sensor system cooperative detection.Consensus behavior means that all agents have the same value of state information or output information under the control algorithm.This thesis mainly studies the event-triggered control and distributed optimization in multi-agent systems to reach consensus.Compared with the energy consumption and resource waste caused by the periodic sampling control,the event-triggered control adopts ape-riodic sampling control,which reduces the number of sampling,and is a very effective control method.Distributed optimization,the optimal solution of the optimization prob-lem is solved by the convex optimization algorithm,that is,the values of all the agents are consistent.It can be seen that event-triggered control studies multi-agent systems from a control perspective,while distributed optimization is studied from an optimization per-spective.Firstly,the event-triggered control based on state feedback is implemented to realize leader tracking consensus control of multi-agent systems.The leader tracking state con-sensus means that there is an autonomous agent in the system as a leader,other agents by the action of the controller,so that their state can finally converge to the leader's state.According to when sampling,when interacting with neighbor information,and when to update the controller,we summarize three different trigger mechanisms,and finally con-verge to the leader's state under event-triggered control.Under three trigger mechanisms,it can exclude Zeno behavior,that is,infinite number of triggering or events can be avoided in finite time.At last,simulation experiments were carried out to verify the effectiveness of the controller.Secondly,the state observer-based event-triggered control is implemented to realize consensus control of multi-agent systems.When the system is a linear multi-agent system,a new trigger condition and relative error vector are designed by using the matrix expo-nential function.This trigger condition is called time-dependent trigger condition.The network topology is a directed graph with a directed spanning tree.Two different observer-based event-triggered controllers are designed to ensure the stability of the closed-loop system and eliminate Zeno behavior.Simulation experiments can verify the correctness of theoretical analysis.Thirdly,the functional observer-based event-triggered control is implemented to re-alize consensus control of multi-agent systems.At begin,under the directed topology of weight-balanced,we propose two different functional observer-based event-triggered controllers to leader tracking consensus control.Then,for the time-dependent trigger con-dition,under the directed topology of the directed spanning tree,the proposed controller makes the system reach the leaderless consensus control.Finally,the effectiveness of the controller can also be demonstrated when it is extended to multi-agent system contain-ment control of multiple leaders.Simulation experiments in each section can show the effectiveness of the controller.Fourthly,distributed optimization problem of multi-agent systems is considered,where a group of agents cooperatively solve a global optimization problem with the objective function as the sum of privately known local objective functions of the agents.For the unconstrained optimization problem,a distributed ADMM optimization algorithm based on consensus is proposed firstly,and then the first-order approximation method is used to propose an inexact consensus ADMM algorithm.Under both algorithms,the optimal algorithm can be achieved.A linear convergence rate is achieved.For the optimization problem with constraints,first convert to the dual problem,and then use the skills of the Laplacian matrix to form a new problem.Design a dual consensus ADMM algorithm,and an improved inexact dual consensus ADMM algorithm to achieve optimality.
Keywords/Search Tags:multi-agent systems, consensus, distributed event-triggered controller, func-tional observer, distributed inexact consensus ADMM algorithm
PDF Full Text Request
Related items