Published in IET Control Theory and Applications
Received on 8th November 2009
Revised on 13th March 2010
doi: 10.1049/iet-cta.2009.0574
ISSN 1751-8644
Brief Paper
Formation control of networked multi-agent
systems
D. Xue
1
J. Yao
1
G. Chen
2
Y.-L. Yu
1
1
Department of Control Science and Engineering, Tongji University, Shanghai, People’s Republic of China
2
Department of Electronic Engineering, City University of Hong Kong, Hong Kong SAR, People’s Republic of China
E-mail: yaojing@tongji.edu.cn
Abstract: In this study, a systematic framework is developed for the consensus control problem, particularly for
formation control o f networked dynamic agents. In view of the complexity of the framework with switching
coupling topology and non-linearity, a new decent ralised formation strategy based on artificial potential
functions (APF) is proposed. Owin g to the existence of local minima in the APF, the formation controller is
designed to introduce some special functions to settle that limitation. A new concept of relative-position-
based formation stability is defined, and a Lyapunov approach is used along with an extended linear matrix
inequalities (LMI) algorithm to analyse the condition for formation stability. Finally, an example with
simulations is provided to demonstrate the effectiveness of the designed formation controller.
1 Introduction
Consensus problems for multiple agents systems (MAS) have
a long history in biological studies [1–3]. Examples in nature
include schools of fish, herds of animals and colonies of
bacteria [2–4]. Referring to the literature, it is obvious that
consensus behaviour has certain advantages, such as
increasing the possibility of finding foods and avoiding
predators. But it requires communications among the group
members and individual decision making based on the
trajectory and velocity of the group.
Advances in communication and computation have enabled
the consensus studies to evolve into the field of engineering
applications, such as coordination of autonomous unmanned
vehicles (AUV), synchronisation of coupled oscillators [5, 6],
navigation of satellites in space and flocking of mobile agents
[4, 7]. Based on some fundamental principles in biological
systems, the motion formation and obstacle avoidance
problems of MAS have attracted considerable attention [5,
8]. This paper only focuses on the formation control of
MAS. In general, there are two main control strategies for
formation: centralised control and decentralised control. The
first one includes the leader-following strategy [9, 10] and
virtual structures, whereas the latter is somewhat similar to
the leader-following approach except that the leader is virtual
and the formation is seen as a rigid body [11].Inthe
centralised control architecture, agents need to have global
information about the whole system. In this framework, the
main problem involves computational complexity and
susceptibility to communication failures [8].Onthe
contrary, decentralised formation mostly relies on the
behaviour of each agent individually, such as collision
avoidance, obstacle avoidance, formation keeping and goal
seeking, based on which global performance is achieved from
a weighted average of the control actions of all agents [12].
The decentralised approach often includes the use of some
artificial potential functions (APF), which have been studied
extensively for path planning of multiple agents in the past
decades, for its mathematical simplicity, ease of
understanding, and good real-time and high-efficiency
performance [4,7,8,13].
A number of recent papers on APF-based controllers for
formation suffer from a common difficulty in dealing with
local minima that often occur in a potential field
2168 IET Control Theory Appl., 2010, Vol. 4, Iss. 10, pp. 2168 – 2176
&
The Institution of Engineering and Technology 2010 doi: 10.1049/iet-cta.2009.0574
www.ietdl.org