52nd IEEE Conference on Decision and Control
December 10-13, 2013. Florence, Italy
Distributed Adaptive Leader-following control for multi-agent
multi-degree manipulators with Finite-Time guarantees
Muhammad Nasiruddin Mahyuddin*, Guido Herrmann** and Frank L. Lewis***
Abstract A robust distributed adaptive leader-following
control for multi-degree-of-freedom (multi-DOF) robot
manipulator-type agents is proposed to guarantee finite-time
convergence for leader-following tracking and parameter
estimation via agent-based estimation and control algorithms.
The dynamics of each manipulator agent system of n degrees
including the leader agent are assumed unknown. For a
specific leader-following network Laplacian, the agents
position, velocity and some switched control information can
be fed back to the communication network. In contrast to the
current multi-agent literature for robotic manipulators, the
proposed approach does not require a priori information of the
leaders joint velocity and acceleration to be available to all
agents due to the use of agent-based robust adaptive control
elements. Due to the multi-DOF character of each agent,
matrix theoretical results related to M-matrix theory used for
multi-agent systems needs to be extended to the multi-degree
context in contrast to recent scalar double integrator results.
A simulation example of two-degree of freedom manipulators
exemplifies the effectiveness of the approach.
I. I NTRODUCTION
Distributed control of multi-agent systems have sparked a
substantial interest due to its significantly broad applications
in many fields such as swarming, flocking, rendezvous and
formation in mobile robots, unmanned aerial vehicles (UAV)
and multi-manipulators. Prominent work shows that consen-
sus control of multi-agent systems involves not just single-
integrator and double-integrator dynamics type systems [1],
[2], [3], [4], [5], [6], [7], [8], [9] but also a group of inter-
connected multiple degree of freedom (multi-DOF) systems
[4], [10], [11], [12], [13]. The leader-following distributed
consensus multi-agent problem for multi-manipulators saves
the computational effort and simplifies the control imple-
mentation [13]. The field of cooperative control of multiple
manipulators has introduced a distributed and cooperative
control structure different from the centralised [14] or a pure
master-slave structure [15].
In this paper, leader-following distributed control of
robotic manipulators of n degrees-of-freedom acting as
agents is considered. In particular, this paper considers finite-
time convergence for synchronization between leader and
follower, but also for parameter adaptation. According to
Wang and Xiao in [16], [5], finite-time consensus allows
better disturbance rejection, enhances robustness against un-
certainties and increase control accuracy [16].
Recent work on cooperative control of multi-manipulator
systems has advanced from scalar [1], [2], [4] to multi-
degree-of-freedom agents [10], [11], [12]. The use of neural-
*Muhammad Nasiruddin Mahyuddin is currently a PhD student in
the Department of Mechanical Engineering, University of Bristol, BS8
1TR, UK. He is also a staff on study-leave at the School of Elec-
trical and Electronics Engineering, Universiti Sains Malaysia. Email:
memnm@bristol.ac.uk or nasiruddin@ieee.org
**Guido Herrmann is a Reader in Dynamics and Control with the
Department of Mechanical Engineering, University of Bristol, BS8 1TR,
UK. Email: g.herrmann@bristol.ac.uk
***Frank L. Lewis is the Moncrief-ODonnell Professor of Electrical
Engineering and Senior Fellow of the Automation and Robotics Re-
search Institute at The University of Texas at Arlington, USA. Email:
lewis@uta.edu
978-1-4673-5717-3/13/$31.00 2013 IEEE 1496
networks [10] to estimate the agents nonlinearities showed
to be beneficial to aid the network consensus, which provides
exponential convergence and ultimate boundedness guaran-
tees of the synchronization error. The work in [11] requires
each agent to know the leaders joint velocity. In contrast,
work in [4] demonstrates the finite-time consensus reaching
of double-integrator systems and multi-robot systems, in
particular, for a leader-following objective. Here the multi-
robot systems are of single DOF in nature and each agent
requires its own and its neighbors mass/inertia parameters,
which simplifies the construction of the control law and
the stability analysis. Another consensus control algorithm
[12] introduces a constant position demand or with the
requirement for enhanced synchronization error information
of not only direct neighbors. This introduces a two-hop
neighbor information. This information is in addition to the
requirement for the leaders initial joint position by all agents
[12].
In contrast to recent work, it is of interest of this paper to
propose a distributed, adaptive, finite-time leader-following
consensus control algorithm for a robotic manipulator multi-
agent system, extending recent results [17] for scalar agents
to the context of multi-degree-of-freedom agents. Leader in-
formation (e.g. velocity) for agents which are not connected
to the leader is avoided. To allow for this extension, matrix
theory [18] usually developed for scalar agent and leader
systems (e.g. [19]) has to be extended to the context of multi-
degree-of-freedom agents. This facilitates the formulation of
a distributed controller for multi-degree-of-freedom systems
and the stability analysis, e.g. the construction of a Lyapunov
function suited to this context. The control law provides
finite-time convergence of the synchronization error and an
adaptive parameter estimation error: This is based on an
extension of [17], where also strong inspiration is taken
from [4]. However, it is to note here that the leader provides
information only to particularly pinned agents, while agents
obtain information only from their neighbors in terms of
position, velocity and a switched control component of the
neighbor.
The next section introduces a generic communication
network concept, necessary to define neighboring agents and
the leader-agent communication.
II. L EADER - AGENT C OMMUNICATION S TRUCTURE
Consider a directed tree G = (V, E) with nonempty
finite set of N nodes V = {v1 , . . . , vi , . . . , vN } where node
i represents the i-th agent. The defined graph is strongly
connected and fixed consisting of directed edges or arcs
E V V with no repeated edges and no self loops
(vi , vi ) 6 E, i. The connectivity matrix is denoted as
A = [aij ] with aij > 0. ThePin-degree matrix is a diagonal
matrix D = [di ] with di = j6=i aij the weighted in-degree
node i (i.e. i-th row sum of A). The graph Laplacian matrix
which is defined as L = D A, L = [lij ], i, j = 1, , N,
has all row sums equal to zero. The connectivity matrix A
and L are irreducible [1], [19]. The leader communication
 T T T T
is again directed from leader (node 0) to agent manipulator where q1 = [q11 , . . . , q1i , . . . , q1N ] RnN and q2 =
only, which is identified through the pinning gain bi 0. [q2i , q21 , . . . , q2N ] R , = [1T , . . . , iT , . . . , N
T T T T nN T T
]
Thus, in case, an agent i, (0 < i N ), is pinned, then nN T
R , = [1 , . . . , i , . . . , N ] T T T lN
R , M =
bi > 0. Thus, bi 6= 0 if and only if there exists an arc from diag([M11 , . . . , Mi1 , . . . , MN 1
]) RnN nN and vg =
the leader node to the i-th node in G. diag([vg1 , . . . , vgi , . . . , vgN ]) RnN N l .
III. M ANIPULATOR DYNAMICS B. Manipulator Dynamics of the leader
A. Agent Dynamics The leader manipulator satisfies the following general
We assume the general structure of the robot dynamics of nonautonomous dynamics in a second order Brunovsky form,
each agent [20] as:
q10 = q20 , q20 = M01 (V0 (q10 , q20 )q20 G0 (q10 )+ 0 ) (9)
Mi (qi )
qi + ci (qi , qi ) + Gi (qi ) = i (1)
where q0 = [q10 q20 ]T Rn is the leaders corresponding
n
where qi = qi (t), qi = qi (t), qi = qi (t) R are the joint position and velocity. It is assumed that the dynamics of
robot arm joint position, velocity and acceleration vectors the leader manipulator remain bounded, i.e. the leader state
respectively; i Rn , the input torque vector of the i-th q0 remains bounded.
manipulator; Mi (qi ) Rnn and Mi (qi ) > 0, is the inertia The leader manipulator dynamics can be regarded as a
matrix, a function of the n joint positions qi , ci (qi , qi ) Rn command generator:
which represents the Coriolis/centrifugal torque, viscous, and Property 3: It is assumed that the reference trajectory
nonlinear damping. Gi (qi ) Rn is the gravity torque vector. of the leader manipulator q0 is at least twice continuously
Several essential properties for (1) facilitate the distributed differentiable with time t and q0 is sufficiently rich (SR) over
adaptive motion synchronisation control system design: any finite interval [t, t + T ] of the specific length T > 0 with
Property 1: The left hand side of (1) can be linearly respect to i (q0 , q0 ), so that
parameterised as such, Z t+T

Ti (q0 (), q0 ())(q0 (), q0 ())d() > I (10)
Mi (qi )
qi + Vi (qi , qi )qi + Gi (qi ) = i (qi , qi , qi )i (2)
t
where i Rl is the system parameter vector containing for some > 0.
l parameters to be estimated, i (qi , qi , qi ) Rnl is The leader-following problem is to design a set of decentral-
the known dynamic regression matrix [22]. The Corrio- ized torque control laws i for the i-th manipulator to drive
lis/centrifugal matrix and the gravity matrix in the left hand each maniputlator to move in synchrony whilst following a
side of (1) can be also linearly parameterised as such, virtual leader, i.e. qi = qj = q0 . The relevant inter-agent
ci (qi , qi ) + Gi (qi ) = Vi (qi , qi )qi + Gi (qi ) = vgi (qi , qi )i communication is specified in the next section.
(3) IV. L EADER - FOLLOWER C ONSENSUS PROTOCOL
Property 2: The inertia matrix Mi (qi ) is symmetric and
positive definite, satisfying the following inequalities: The control protocol proposed here is for the case of multi-
agent MIMO systems instead of multi-agent SISO systems as
c1 kk2 T Mi (qi ) c2 kk2 , Rm , (4) in the authors previous work in [17]. To solve this particu-
lar leader-following consensus problem, the synchronisation
where c1 and c2 are known positive constants. errors (position and velocity) for the i-th agent are defined
The regression matrix, i is given in Property 1. It has as N
the acceleration as argument. Note that in our proposed X
adaptive control algorithm, the regression matrix will not eAi = aij (q1j q1i ) + bi (q10 q1i ) (11)
use joint acceleration unlike in [23]. This is inspired by i=1
[20] where similar approaches are used to avoid acceleration N
X
measurements. eBi = aij (q2j q2i ) + bi (q20 q2i ) (12)
Denote the following states for joint variables for each i=1
agent manipulator,
qi = q1i , qi = q1i = q2i (5) The synchronization errors (11)-(12) are influenced only
by their corresponding direct neighbours dynamics whose
where q1i Rn and q2i Rn are the agent manipulators connections depend on the graph description of L. This
joint position and velocity respectively. Then, express the error (11)-(12) and a later introduced bounded switched term
agent manipulator dynamics in (1) in a Brunovsky form, specific to an agent perceived by its direct neighbour can
only be used by a particular agent for control purposes. The
q1i = q2i , q2i = Mi1 (Vi (q1i , q2i )q2i Gi (q1i ) + i ) (6) consensus error in (11) and (12) can be also expressed in
By the linearity-in-the-parameter assumption as stated in terms of the overall network as
Property 1, (6) can be expressed as, EA = [(L + B) In ] (q1 q10 ) (13)
q1i = q2i , q2i = Mi1 (vgi i + i ) (7) EB = [(L + B) In ] (q2 q20 ) (14)

where i is the agent manipulators parameters associated where L + B RN N describes the communication
with the Coriolis/centrifugal and gravity matrix to be esti- topology of the leader-following multi-agent network. The
mated by the novel parameter estimation algorithm presented pinning gains are B = diag(b1 , , bi , , bN ) RN N ,
in this note. In Rnn is the identity matrix, qh0 = 1N q0 RnN , h
The overall agent manipulator dynamics can be expressed {1, 2} (Noting that 1N = (1, 1, , 1N )T RN ).
as, Property 4: The communication topology L can be
+


q1 = q2 , q2 = M vg (8) framed so that the irreducible L is upper triangular.

1497
This property avoids loops but also introduces a specific B. Auxiliary Integrated Regressors
leader-agent structure. The filtered torque formulation is now considered for an
Suppose h = qh qh0 , (h {1, 2}) represents the auxiliary regressor used for the adaptation algorithm. Define
disagreement vector to be used only for analysis. Then, the a filtered regressor matrix Wi (t) and vector Ni (t) as,
synchronisation error vector Ek = [ek1 , eki , . . . , ekN ]T
R2N , k {A, B}, i is assumed to be bounded by W i (t) = kF Fi Wi (t) + kF Fi Tfi (q1i , q2i )fi (q1i , q2i ),
kh k kek k/((L + B) In ), k {A, B},
h {1, 2} Wi (0) = wIl , (25)
(15) N i (t) = kF Fi Ni (t) + kF Fi Tfi (q1i , q2i )fi , (26)
where () denotes the minimum singular value of a matrix
and e = 0 if and only if the nodes synchronise, i.e. Ni (0) = 0

q1i (t) = q0 (t), i = 1, . . . , N (16) where, kF Fi R+ , can be interpreted as a forgetting factor.

The solution of Wi (t) (25) shows that Wi (t) wIl ekF Fi t
The errors (11)-(12) are utilized in the novel adaptive law for w > 0. This bound will be exploited in the Lyapunov
proposed in this note: Through the inclusion of auxiliary fil- analysis section later. Having formulated the auxiliary torque
ters in each of the agent manipulator system, it can be shown filters and filtered regressors, (24) can be expressed in an
that finite-time convergence of the leader-following synchro- overall expression for the network as
nization error and finite-time adaptation can be achieved. The = f
next section focusses on the introduction of an adaptive law f (q1 , q2 ) (27)
as used by each agent. where f (q1 , q2 ) = diag(f1 (q11 , q21 ), . . . , fi (q1i , q2i ), . . . ,
V. F INITE -T IME PARAMETER E STIMATION A LGORITHM fN (q1N , q2N )) RN (nl) .
=
T T T T Nl
A. Auxiliary Torque Filters [1 , . . . , i , . . . , N ] R . Moreover,
In this section, an auxiliary filtered regression matrix (t) = W
N (t)
(IN ekF F t wIl )
(28)
and suitable filtered vectors for the adaptation algorithm
will be formulated for each agent manipulator, based on where N (t) =
= [N1T , N T , . . . , N T ]T RN l , and W
i N
N lN l
its torque measurement i . By having the torque measure- diag(W1 , Wi , . . . , WN ) R .
ment filtered, acceleration measurements for the regressor
i (q1i , qi , qi ) can be avoided [20], [21]. Indeed, the regressor C. Parameter Estimation Laws
i (q1i , q2i , qi ) in (2) uses joint accelerations which generally The parameter estimation algorithm comprises of a
is not practical. Hence, the equation (1) can be written as, switched parameter Ri for each agent manipulator i:
i = fi + hi (17) i = i Ri
(29)
The components of torque can be split and defined as,
Wi (t) i N (t)i
d Ri = 1i i N (t)i ),
+ 2i (Wi (t)
fi = [Mi (q1i )q2i ] (18) kWi (t) i Ni (t)k
dt
hi = M i (q1i )q2i + Vi (q1i , q2i )q2i + Gi (q1i ) (19) i = 1, . . . , N. (30)
= h1i + h2i (20) where 1i and 2i are positive scalars which are to be chosen
i (q1i )q2i and h2i = Vi (q1i , q2i )q2i + large enough in the Lyapunov based design to achieve robust
where h1i = M stability. i is a diagonal positive definite matrix.
Gi (q1i ). By virtue of the linearity-in-the-parameter assump- Remark 1: In [17], it has been shown that (10) implies
tion, the split terms can be parameterised as such, that W0 (t) is invertible with well defined bounds for the
fi = Mi (q1i )q2i = m1i (q1i , q2i )i (21) smallest and largest singular value.
The agent-specific adaptive law (29) will be now used as
i (q1i )q2i = m2i (q1i , q2i )i
h1i = M (22) part of a distributed control law for each agent.
h2i = Vi (q1i , q2i )q2i + Gi (q1i ) = vgi (q1i , q2i )i (23) VI. D ISTRIBUTED A DAPTIVE C ONTROL L AW
The concept of robust sliding mode control for a finite time
Filtering the terms m1i , m2i , vgi and through an im-
sliding plane is introduced to allow for finite-time conver-
pulse response filter f = 1 e1/kt to produce m1f i = gence of the synchronisation error. The approach presented
f m1i , m2f i = f m2i , vgf i = f vgi and f = f here is suitably combined with an adaptive control element
respectively. The filtered computed-torque equation can be to enhance consensus control performance by incorporating
rewritten as, finite-time parameter estimation.
m1i (q1i , q2i ) m1fi (q1i , q2i )
[ + m2fi (q1i , q2i ) A. Sliding variable Definition
i
Note that EA = EB . Denote m as the index for one of n
+vgfi (q1i , q2i )]i = fi (24) joints. Thus, the sliding variable rim is defined for each joint
fi (q1i , q2i )i = fi m, m {1, 2, . . . , n} of agent manipulator i,
where fi (q1i , q2i ) Rnl , i Rl . By comparison to rim = |eBim | sign(eBim ) + im eAim R1 (31)
(1), the filtered system equation of (24) clearly avoids the
T n
acceleration measurements which are sometimes practically where ri = [ri1 , , rim , , rin ] R is the sliding
unavailable. Note that i (q1i , q2i , q2i ) is the unfiltered re- error for agent manipulator i. The scalar satifies 1 < <
gressor for fi (q1i , q2i ). 2 and m > 0. It can be shown that the sliding variable

1498
rim leads to finite time convergence of the closed-loop, i.e. where the auxiliary torque input ci is:
rim = 0 is governed by
1/
e Aim = i |eAi |1/ sign(eAi ), (32) 
1 N
 X
ri

ci = aij cj + ki
The sliding variable ri for each manipulator i is therefore (di + bi )
j=1 k
r i k

ri = Bi + eAi Rn (33) j 6= i
where Bi = [|eBi1 | sign(eBi1 ), |eBi2 | sign(eBi2 ), (42)
. . . , |eBin | sign(eBin )]T

Rn and =
and ri = Ei ri . The control gains i are chosen so that
diag(i1 , i2 , . . . , in ). The sliding variable in (33) the matrix diag(1 M11 , 2 M21 , , N MN 1
) satisfies the
can be expressed for the overall network, conditions (38) for matrix N . This implies for suitable choice
B ) + (IN )EA RnN of ki > 0 that the parameter estimation errors Wi (t) i
r = E(E (34)
N (t)i , (i = 1, , N ) and synchronisation errors EA and EB
B ) = [T , T , . . . , T ]T . Differentiating r
where E(E converge to 0 in finite time in an arbitrarily large compact
B1 Bi BN
yields, set of EA , EB and i determined by k, 1i and 2i . The
EB + (IN )EA
r = E (35) parameter estimates converge to its true values.
The overall combined control law can be written as
where + +
= vg c r (43)
= diag(E
E 1 , . . . , E
i , . . . , E
N ) R(N nN n) (36)
where = In with = diag(1 , i , . . . , N ) RN N
i defined as
with E and = [
T1 ,
T,...,
T ]T . The corresponding auxiliary
i N
i = diag(|eB1 |1 , |eB2 |1 , , |eBn |1 ) Rnn torque input which encapsulates the decentralized switching
E laws for all agents is:
(37)
1
B. Leader following control law c = [[(D + B) In ]] [(A In )c + KSIGN (
r)] (44)
A set of adaptive control laws are to be defined in where K = diag([k1 , ki , . . . , kN ]) and
this section, which will solve the leader following control  T
problem within a finite time. To facilitate the analysis and r1 ri rN
SIGN (r) = , ,..., (45)
design a result known from the cooperative control literature, kr1 k k ri k k
rN k
e.g. [19], is extended to the context of multi-degree-of- h i
1
freedom systems. Note that I + [(D + B) In ] (A In ) =
Lemma 1: Let L RN N be an irreducible and upper 1
[(D + B) In ] [[(D + B) In ] + (A In )]. Invoking
triangular matrix and B RN N may have at least
the associative property of the Kronecker product, i.e.
one diagonal element. Moreover, there is a matrix N = A F + B F = (A + B) F and since L = D A,
diag(N1 , N2 , , NN ) RnN nN for which Ni Rnn , then (44) can be simplified as,
i = 1, , N , are positive definite and the following in-
equalities hold c = [[(L + B) In ]]1 [KSIGN (
r)] (46)
(Ni ) > kNi+1 k > 0, i = 1, , N 1, (38) Proof of Theorem 1: The following Lyapunov function is
proposed,
then there exists a matrix P
V = Vr + V (47)
P = P P In , P = diag(x1 /y1 , x2 /y2 , , xN /yN ), 1 T 1 T 1 1 1
= r P r + N W W N (48)
x = (L + B)1 1N , y = (L + B)T 1N , 2 2
kNi k is defined as
1N = [1, 1, , 1]T , 1N RN , max > maxi=1,...,N where N
(Ni )
(t) = N
N (t) W (t) (49)
P = diag(1, max , , (N 1)
max ), (39)


= W (t) KIef f , (50)
so that:
KIef f = (IN ekF F t w) RN ll , =
,
P ((L + B) In )N + (((L + B) In )N )T P > 0 (40) (t) = diag(W1 (t), W2 (t), . . . , WN (t)) and N (t) =
W
T T T T
[N1 , N2 , , NN ] . Note that,
The proof of this Lemma can be found in the appendix. The
diagonal matrix P from Lemma 1 will be used in the leader (t) ekF F t wIl W (t) ekF F t 1 . (51)
1
W
following control law in the following Theorem: w

Theorem 1: Consider the multi-manipulator system with To compute Vr in our analysis, (36) is used to denote the
dynamics defined by (1), adaptive parameter estimation al- following:
gorithms (29) and communication interconnections between r = E r (52)
manipulator agents and its corresponding virtual leader
defined through the given Laplacian matrix L. The adaptive Differentiating (13),
control law i , for each agent manipulator i is: EA = [(L + B) In ] (q2 q20 ) (53)
i = vgi i ri ,
i + i ci + i E (41) EB = [(L + B) In ] (q2 q20 ) (54)

1499
The derivative of the sliding mode term in (35) can be written where for 2 = diag((Il 21 ), (Il 2i ), . . . , (Il 2N )),
as 2 > 0 the matrix is:
[(L + B) In ] M vg + Q B
0 + E Q P [(L + B) In ]Mvg W 1
 
r = E



= 1 (60)
(55) 2 W
where Q 0 = 1N [M 1 (V0 (q10 , q20 )q20 G0 (q10 ) + 0 )]
0 This leads after some manipulation to
and = (IN ). Differentiating V yields,
V (E) 2 Vr + V


1 1 W p p
h i
V = rT P r + N TW 1 N (56) N Vr N V ,
(E) (61)
t
Computing the derivative of W 1 N
=
W
1 KIef f for suitable choice of gains ki and i so that > 0 exists.
provides Following arguments of [4] and [17], this guarantees that
h 1 all agent trajectories follow the persistently exciting demand
+
i
W N = (57)
t trajectory of the leader by Property 3 within finite time, so
that (Wi1 ) and (Wi ) remain finite and strictly larger than
where KkIef hf = kiF F KIef f and = converge to their true values.
1 KIef f kF F W W
1 = [1 , , N ] for Rn . 0. The estimates, 
W
def def
VII. L EADER -F OLLOWING M ULTI -M ANIPULATOR
We may now define ri = Ei ri , EB = E 1 EB , E XAMPLE

i = + 2i i e kF F t
wi , = [1 , , N ] and A simulation example is presented to illustrate the perfor-
def mance of the proposed distributed adaptive leader-following
= [(L + B) In ]Q 0 + EB + W
1 KIef f ,
def control algorithm: we consider a simple network of two
= [1 , . . . , N ] . Note that EB is not singular due to
T
manipulators of 2 DOF (revolute planar) with a leader
the choice of , 1 < < 2. Exploiting the fact that P and whose communication topology is defined by a Laplacian
E are diagonal, adopting the control torque in (43) and 
1 1

incorporating the auxiliary torque input c (44) and (50) it matrix, L = 0 0 . The diagonal pinning matrix is
follows: B =diag(0, 1). The manipulator leader is controlled by
1 T means of a feedback linearisation controller following sinu-
V = r (P [(L + B) In ][M ]
2 soidal/SR signals for both joints. Table I shows the masses
+[M]T [(L + B) In ]T P ) r of the manipulator agents links to be estimated.
T
r P [(L + B) In ]Mvg W
+ 1 N
TABLE I
M ANIPULATOR S YSTEM PARAMETERS ,mi = [m1i , m2i ]T
rT P [(L + B) In ]M
[(L + B) In ]1
N N
Manipulator i Link 1 mass (kg) Link 2 mass (kg)
riT Pi i + T W 1 1 i
X X
[KSIGN (
r)] + N i i i Leader 2.35 3
i=1 i=1 Agent 1 1 1.35
N N Agent 2 0.5 1

T W 1 Ni T W 1 N
i
X X
1i N i i 2i N i i (58)
ik
kN
i=1 i=1 8
Agent 2
20
Agent 2
We may now analyse the matrix 6 Agent 1 Agent 1
15
4 Leader Leader
= rT P [(L + B) In ]M
Joint 1 Position( )

[(L + B) In ]1 K
Joint 2 Position( )

Setpoint Setpoint
M (59) 2
10

Note that (L + B) is upper triangular, i.e. its inverse is 0 5

also upper triangular. Thus, the structure of M = [M ij ], 2 0

(M ij Rnn ) follows also an upper triangular structure, 4
5
i.e. M ij = 0 for i > j. Note also the diagonal structure of P 6
10
(39) and the symmetry of Mi1 , which implies that M ii are 8

all symmetric. The inverse matrix of (L+B) can contain only 0 2 4 6

(a) time(s)
8 10 0 5 10
(b) time(s)
15
non-negative elements. Thus, the definition of M implies
that M ii is also positive definite. We may now write the Lya-
punov matrix P (39) as P = diag(P1 In , P2 In , . . . , PN In ). Fig. 1. Joint 1 and Joint 2 Position for for agent i and their leader with
proposed finite-time distributed adaptive control.
Employing now Lemma 1, it follows the following upper
bound for V: Figure 1 shows all the agents joint position trajectories
T for joint 1 and joint 2 with different initial conditions. The
r r leader tracking of all the agents is observed to be finite-
V time. All the agents successfully follow the leader within
N N
less than 1 sec. Thus, the leader-following task by each
N

N
agent has been accomplished. Figure 2 shows the finite-time
X
ii )
X
ij k Pi i
convergence of the respective agents link masses estimates
k ri k ki (M kj kM within less than 1.5 seconds. The exemplified result shows
i=1 j=i+1 that the local finite-time parameter estimation algorithm by
N   each agent also enhances the convergence of the network
i k 1i (W 1 ) 1 1
X
kN i Wi i i consensus in addition to the switching signal fed back to the
i=1 network.

1500
 1.5
2 for Agent 2 and Q = P ((L + B) In )N , Q = [Q ij ], Q
ij Rnn .
1.35 kg
The choice of P and the diagonal dominance
P of P (L + B)
ii ) > i1
1
1 kg (and P P (L+B)) guarantees (Q j=1 Q ji , where

Qji = 0 for j < i. This implies that Q = Q + QT satisfies
1 for Agent 1, m

0.5 kg the conditions of Lemma 2, i.e. the symmetric matrix Q is

0.5
positive definite.
1
m1 1
m2 2
m1 2
m2 R EFERENCES
0
m

[1] A. Das and F. L. Lewis, Cooperative adaptive control for syn-

0 0.2 0.4 0.6
Fig. 2. Link Mass Estimation for for agent i {1, 2}
VIII. C ONCLUSIONS
A novel distributed leader-following adaptive controller
of multi-manipulators has been presented. It is shown that
the proposed approach guarantees finite-time convergence
for leader-following tracking and parameter estimation via
agent-based estimation and control algorithms. It is shown
that the extended matrix theoretical results related to M-
matrix theory used for multi-agent systems is instrumental
to the multi-degree context. This allows to prove the exis-
tence of a Lyapunov function which is used in the analysis
for stability and finite-time convergence of the consensus
error and parameter estimation error. Information on the
leaders dynamics is only required by pinned agents and
the dynamic interaction by all agents is fully defined by
the communication network: Any unknown dynamics are
compensated by the switching control which is fed-back to
the communication network and therefore, the leaders joint
position and velocity are not required a priori. Numerical
examples of a two-degree of freedom two-agent system with
one leader prove the feasibility of the results.
A PPENDIX 1: P ROOF OF L EMMA 1
To facilitate the proof of Lemma 1, the following prelim-
inary Lemma is a necessary extension of work for instance
found in [18].
Lemma 2: Suppose Q RnN nN is defined as below:
Q11 Q12 . . . Q1N
Q21 Q22 Q2N
Q= . . ..
. .
QN 1
nn
where Qij R , i, j = 1,
satisfying:
P
N trol of Dynamical Systems. Springer Verlag, Heidelberg, 2009.
Qii > 0, Qij,i6=j 0, (Qii ) >
j=1,i6=j Qij,i6=j
(62)
then (i) the matrix Q is invertible and (ii) any real eigenvalue
of Q is positive.
From [19, Chapter 4, p.174] follows for P , (L + B) that
P (L + B) is upper triangular and diagonally dominant in
terms of row and column vectors, i.e. P (L + B) + (L +
B)T P > 0 is positive definite. From max > 1 and the
definition of P (39), this easily also implies that P P (L +
B) is again diagonally dominant in terms of row and column
vectors so that P P (L + B) + (L + B)T P P > 0. Now, we
may investigate
Q = P ((L + B) In )N + (((L + B) In )N )T P (63)
From (38) and the diagonal dominance
of the upper
trian-
PN
gular P (L + B), it follows (Qii ) >
j=i+1 Qij
for
0.8 1 1.2 1.4 1.6 1.8 2
time(s) chronization of second-order systems with unknown nonlinearities,
International Journal of Robust and Nonlinear Control, 2010.
[2] , Distributed adaptive control for synchronization of unknown
nonlinear networked systems , Automatica, 2010.
[3] R. Olfati-Saber, J. A. Fax, and R. M. Murray, Consensus and
Cooperation in Networked Multi-Agent Systems, Proceedings of the
IEEE, vol. 95, no. 1, pp. 215233, Jan. 2007.
[4] S. Khoo, L. Xie, and Z. Man, Robust finite-time consensus tracking
algorithm for multirobot systems, Mechatronics, IEEE/ASME Trans-
actions on, vol. 14, no. 2, pp. 219 228, april 2009.
[5] F. Xiao, L. Wang, and J. Chen, General distributed protocols for
finite-time consensus of multi-agent systems, in CDC, 2009 held
jointly with the 2009 28th CCC. CDC/CCC 2009. Proceedings of the
48th IEEE Conference on, dec. 2009.
[6] J. Fax and R. Murray, Information flow and cooperative control of
vehicle formations, Automatic Control, IEEE Transactions on, vol. 49,
no. 9, pp. 1465 1476, sept. 2004.
[7] R. Beard and E. Atkins, A survey of consensus problems in
multi-agent coordination, Proceedings of the 2005, ACC, 2005..
[8] A. Jadbabaie, J. Lin, and A. Morse, Coordination of groups of mobile
autonomous agents using nearest neighbor rules, Automatic Control,
IEEE Transactions on, vol. 48, no. 6, june 2003.
[9] Q. Hui, W. Haddad, and S. Bhat, Finite-time semistability and
consensus for nonlinear dynamical networks, Automatic Control,
IEEE Transactions on, vol. 53, no. 8, pp. 1887 1900, sept. 2008.
[10] G. Chen and F. Lewis, Distributed adaptive controller design for
the unknown networked lagrangian systems, in Decision and Control
(CDC), 2010 49th IEEE Conference on, Dec.
[11] L. Cheng, Z.-g. Hou, and M. Tan, Decentralized adaptive consensus
control for multi-manipulator system with uncertain dynamics, 2008
IEEE Int. Conf. on Systems, Man and Cybernetics, Oct. 2008.
[12] J. Mei, W. Ren, and G. Ma, Distributed coordinated tracking with
a dynamic leader for multiple euler-lagrange systems, Automatic
Control, IEEE Transactions on, vol. 56, no. 6, pp. 14151421, June.
[13] W. Ren and R. Beard, Consensus seeking in multiagent systems un-
der dynamically changing interaction topologies, Automatic Control,
IEEE Transactions on, vol. 50, no. 5, may 2005.
[14] N. Xi, T. Tarn, and A. Bejczy, Intelligent planning and control
for multirobot coordination: An event-based approach, Robotics and
Automation, IEEE Transactions on, vol. 12, no. 3, pp. 439452, Jun.
[15] Y. Zheng and J. Y. S. Luh, Optimal load distribution for two industrial
robots handling a single object, in ICRA, Proceedings., Apr, 1988.
[16] L. Wang and F. Xiao, Finite-time consensus problems for networks
of dynamic agents, IEEE TAC, vol. 55, no. 4, April 2010.
[17] M. N. Mahyuddin, G. Herrmann, J. Na, and F. L. Lewis, Finite-
time adaptive distributed control for double integrator leader-agent
synchronisation, in Intelligent Control (ISIC), 2012 IEEE Interna-
QN 2 QN N tional Symposium on, oct. 2012, pp. 714 720.
[18] M. Fiedler and V. Ptak, On matrices with non-positive off-diagonal
, n, are symmetric matrices elements and positive principal minors, Czechoslovak Mathematical
Journal, vol. 12, no. 3, pp. 382400, 1962.
[19] Z. Qu, Cooperative Control of Dynamical Systems Cooperative Con-
[20] F. L. Lewis, D. M.Dawson and C. T. Abdallah, Robot Manipulator
Control Theory and Practice, 2nd ed. Marcel Dekker, 2004.
[21] M. Mahyuddin, G. Herrmann, and S. Khan, A novel adaptive control
algorithm in application to a humanoid robot arm, in Advances
in Autonomous Robotics, ser. Lecture Notes in Computer Science,
Springer Berlin / Heidelberg, 2012, vol. 7429, pp. 2536.
[22] M. W. Spong and M. Vidyasagar, Robot Dynamics and Control.
Wiley, 1989.
[23] G. Feng, A new adaptive control algorithm for robot manipulators in
task space, IEEE Transactions on Robotics and Automation, vol. 11,
no. 3, pp. 457462, June 1995.

1501
 IEEE CDC 2013, FLORENCE, ITALY

DISTRIBUTED ADAPTIVE LEADER-FOLLOWING

CONTROL FOR MULTI-AGENT MULTI-DEGREE
MANIPULATORS WITH FINITE-TIME GUARANTEES

1 ,2
MUHAMMAD NASIRUDDIN MAHYUDDIN
GUIDO HERRMANN (SUPERVISOR)1
FRANK L. LEWIS3

BRISTOL ROBOTIC LABORATORY, MECHANICAL DEPARTMENT, UNIVERSITY OF BRISTOL1

School of Electrical and Electronics Engineering, UNIVERSITI SAINS MALAYSIA 2
3
Automation and Robotics Research Institute, UNIVERSITY OF TEXAS

1
 Presentation layout
Motivation
Problem Formulation
Distributed Adaptive Control Algorithm
Analysis
Simulation Results
Conclusions

2
 Acknowledgement
I would like to thank Universiti Sains Malaysia for their support in terms of
sponsorship.

23