Files in this item

FilesDescriptionFormat

application/pdf

application/pdfCHEN-DISSERTATION-2018.pdf (2MB)
(no description provided)PDF

Description

Title:Control and collective intelligence of multi-agent system
Author(s):Chen, Cheng
Director of Research:Baryshnikov, Yuliy
Doctoral Committee Chair(s):Hovakimyan, Naira
Doctoral Committee Member(s):Liberzon, Daniel; Belabbas, Mohamed Ali
Department / Program:Mechanical Sci & Engineering
Discipline:Mechanical Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Control
Multi-agent System
Cyclic Pursuit
Swarm
Covering
Topological Configuration Space
Abstract:Distributed artificial intelligence (DAI) and multi-agent system (MAS) has recently gained increasing interest due to its vast applications in real-world problems. Inspired by the natural MAS, this thesis primarily focuses on the study of the collective intelligence and the joint behavior of MAS, which are typically generated by a group of intelligent agents applied with autonomous controls. In particular, the collective intelligence is described as geometric group patterns, stationary distribution and cooperative motions in this thesis. As the size of the group increases, it is essential to exploit simple and distributed controls to achieve the desired collective intelligence of the system with robustness and minimal cost. Therefore, this thesis proposes two bio-inspired applications of MAS to illustrate that simple controls can obtain stable and robust limiting collective behaviors. Besides, topological configuration space is introduced to describe the admissible collective behaviors and design the cooperative controls. In the first part of the thesis, cyclic pursuit as a periodic joint behavior of the MAS is studied. With prescribed deployments and controls of the agents, the limiting group geometric formation varies from regular polygons to an eight-shaped graph. The rotation number of the graph is a geometric invariant during the evolution. In the second application, the cyclic collective intelligence is generalized to a swarm concentration problem with desired gathering and drifting group behaviors. Randomized algorithms are used to localize the agents and generate stationary distributions of the swarm. In the last part of the thesis, topological configuration space is implemented to assist the design of hybrid controls for a multi-agent coverage problem.
Issue Date:2018-03-28
Type:Text
URI:http://hdl.handle.net/2142/101132
Rights Information:Copyright 2018 Cheng Chen
Date Available in IDEALS:2018-09-04
2020-09-05
Date Deposited:2018-05


This item appears in the following Collection(s)

Item Statistics