www.ietdl.org
Published in IET Control Theory and Applications
Received on 30th May 2013
Revised on 10th May 2014
Accepted on 17th September 2014
doi: 10.1049/iet-cta.2013.1104
ISSN 1751-8644
Brief Paper
Distributed reduced-order observer-based approach to
consensus problems for linear multi-agent systems
Lixin Gao
1
, Bingbing Xu
1
, Junwei Li
1
, Hui Zhang
2
1
Institute of Intelligent Systems and Decision, Wenzhou University, Zhejiang 325035, People’s Republic of China
2
Department of Control Science and Engineering, Zhejiang University, Hangzhou 310027, People’s Republic of China
E-mail: lxgao@wzu.edu.cn
Abstract: In this study, the authors probe the multi-agent consensus problem with general linear dynamics via the distributed
reduced-order observer-based protocols under directed switching topology. A new type reduced-order observer is adopted to
modify the existed consensus protocols. Based on graph theory, Riccati equation and Lyapunov method, a multi-step algorithm
is provided to design the gain matrices involved in the used protocols. The consensus stability problem is investigated by use
of the piecewise Lyapunov functions incorporated with an average dwell time approach. It is shown that if the average dwell
time is chosen sufficiently large and all directed switching topologies have a directed spanning tree, then the multi-agent
system can achieve consensus via the author’s modified protocols. Furthermore, as the special cases, the consensus conditions
under fixed topology and balanced switching topology are obtained easily. To obtain a desired consensus decay rate, the
related gain matrix design approach is provided. Finally, a simulation example is given to illustrate the effectiveness of their
obtained results.
1 Introduction
Recently, many researchers focus on the coordination con-
trol of the multi-agent systems because of its useful applica-
tions in many areas. The consensus control problem is a key
technology in cooperative control, whose aim is to find the
control law that enables a group of agents to reach an agree-
ment on certain quantities [1, 2]. There are many interesting
issues of cooperative control associated with consensus, such
as consensus filtering [3], synchronisation [4], swarm sta-
bility [5, 6], formation [7], network with time-delay [8],
finite-time consensus [9] and so on.
It is well-known that the agent’s dynamics and the multi-
agent interaction topology are two key factors to achieve
consensus. The stochastic matrix and Lyapunov-based meth-
ods are often chosen to analyse first-order [1, 10–12]
and second-order consensus problems [13–16], respectively.
However, it is very interesting to investigate the multi-agent
consensus problem with general linear dynamics [17–20].
Compared with the switching interacting topology, it is rel-
atively easy to handle fixed topology by using the eigenvalue
decomposition approach [20, 21]. As for the switching inter-
action topology case, the common Lyapunov function (CLF)
method is a good substituted way to analyse consensus [13,
16, 22–24]. In [25], the authors expanded the concept of
dwell time to average dwell time and proved that switching
among stable linear systems results in a stable system pro-
vided that the switching law has large enough average dwell
time. The stability problems of linear switched systems
consisting of both Hurwitz stable and unstable subsystems
are investigated via an average dwell time approach in [26].
For some practical systems, an agent cannot obtain full
state information but only output information of its neigh-
bours, which may affect the system to achieve state consen-
sus. To achieve the control goal, a common approach is that
an observer is adopted for the agent to estimate those unmea-
sured variables. To estimate unmeasured velocity of the
active leader, Hong et al. [13] proposed a neighbour-based
estimation rule together with a neighbour-based feedback
law for each first-order follower-agent. Base on [13, 22]
proposed a neighbor-based consensus control law by assum-
ing that each agent can use the neighbour’s estimation
values. In [27], the authors proposed the observer-based con-
sensus algorithm for the second-order agents under input
constraints. The distributed observer-based control protocol
was established for multiple non-holonomic mobile agents
in [28]. To track the active leader with unmeasured veloc-
ity, Hong et al. [15] proposed a distributed observer-based
control law for the second-order follower-agents. In [24],
the authors proposed a distributed consensus protocol based
on an observer to estimate the acceleration of the leader.
In [29], the authors proposed a distributed observer-based
control protocol for each first-order follower-agent to track
the active leader with general linear dynamics and Gao
et al. [30] extended the results of [29] to communication
delay case. A unified framework was introduced in [19]
to address the multi-agent consensus, which proposed an
observer-type consensus protocol using only the relative
IET Control Theory Appl., pp. 1–9 1
doi: 10.1049/iet-cta.2013.1104 © The Institution of Engineering and Technology 2014