Artificial Formation

of

Stable Aggregation
Multi-Agent System
Forces for

Samitha W Ekanayake and Pubudu N Pathirana

School of Engineering and Technology
Deakin University
Geelong, Victoria 3217
Australia
[swek, pubudu] @deakin.edu.au

Abstract- This paper introduces, a robust and stable

algorithm based on artificial formation forces, for Multi-Agent
System (MAS) aggregation in 2D space. The MAS model with
artificial forces; consists of inter-member collision avoidance
element, formation generation element and a velocity based
damping element; is analysed for stability and convergence.
Computer simulations are used to illustrate stability and
convergence, and to demonstrate effectiveness of the algorithm.
Keywords-Aggregation, Formation, Multi-agent System,
Swarm, Stability analysis
I. INTRODUCTION

Corporative robotics, where multiple robots perform collaborative tasks in decentralized manner, has been a research
interest for robotics researchers across the world. Distinct
advantages of MAS, such as robustness to failures of its
members, flexibility in motion, adaptive behavior and low
cost of individual members, compared with single robot based
systems, make them suitable for vast range of applications
(i.e. Nano-robotics for medical applications, Defense , Search
& Rescue etc,). Natural swarms or social foraging insects;
such as bees, ant colonies etc, sets the conceptual framework
for MAS. Researches and studies on group behavior of such
animals/ insects [1], [2], [3], [4], mathematical modeling of
group behavior [5], [6] etc, forms the foundations for MAS
research.
Formation /aggregation of agents for coordinated task is among
fundamental behaviors of MAS, as well as in biological
swarms [1], [3]. Many researchers have performed various
methods for aggregation of MAS. Gazi and Passino introduced
biologically inspired individual attraction/ repulsion based
swarm aggregation model in [7], [8], [9], with stability analysis
and defining a time and physical bounds for convergence.
Behavior based, which is more nature oriented, swarm formation and obstacle avoiding method [10] was practically
implemented and tested in real-world, while Soyal and Sahin
[11] introduces a aggregation method using simple reactive
behaviors. Study by Nuruse etal [12] is different approach in
behavior based aggregation, where emotion and affection like
status determines the aggregation behavior of agents.
In this paper we present a swarm model for multi-agent
society that will converge around a predefined location in 2D

1-4244-0555-6/06/$20.00 ( 2006 IEEE

space. The model assumes that the agents as point objects

with no physical dimensions and having same properties in
each agent (such as mass, mobility, etc,.). The formation
algorithm uses inter-individual repulsion forces in order to
avoid collisions between agents, global formation forces which
converges the swarm around a predetermined location and
a friction-like damping force that brings swarm to a halt.
Also in this model, disturbances, sensor errors, limitations
in control model of the mobile agent and other practical
difficulties were not considered, while assuming that each
agent is having its own localization capabilities. With recent
advances in communication, computing and networking, the
assumption of knowledge ofpositions without error become a
practical statement.
An outline of the paper is as follows: Formation algorithm
together with the swarm model is introduced in section II.
Stability and behavior analysis of the model were carried out
in sections III and IV. In section V we present computer
simulations, to demonstrate the effectiveness of the aggregation
algorithm and to illustrate results obtained in section III.

II. DYNAMIC FORMATION ALGORITHM

Consider a multi-agent system or swarm consisting of N
members in two dimensional euclidean space. The state of the
member i is described by
xiz Zi

Xt =l

where zi, represents the position vector of jth member in 2D

space. (We assume that each agent knows positions of all the
members in the swarm with no time delay.)
Before stating the aggregation algorithm we define
a = ( 1 0 ), Q = ( 0 1 ). Also, we use R for
real number and R+ to represent positive real number.
of
The
state
the
whole
swarm,
( Xl X2 X3
x
.XN )T is determined in
continuous time dynamic model described by,

Page 1 29

A*x+B*u

(1)
ICIA 2006

Where,
A 0 0 ...
0 A 0 ...
0 0 A ...

Where, kfriction C R+ is a constant which determines the

power of the damping action. An analysis of kfriction, with

0
0
0

respect to stability, is presented in section III.

III. BEHAVIOR OF MULTI-AGENT SYSTEM

L

0 0

...

[ B B B ...

2 NxN
T

(2)

and

A = 0 i (1) c B

(3)

of,

The control function u in (1) consists of,

'U = ( Ul zU2 'U3 ..UN

att

)T1

(F tt + Friobot

f riction)

kmass

Ff rictionr(Zi)

FriObot

and

Fiobot(x) respectively.)

Fa in (4) is the attraction force on jth agent from the shape,

defined as
(5)
Fatt =katt (Zd-aXi)
where Zd is the destination/target vector and katt C R+, is a
constant determining the power of attraction force.
Term FiobOt in (4), refers to the force acting on 'th robot due
to the location of other members of the swarm, is defined as
Fr0bOt = krobot

CVX-

(6)

Term Fjriction in (2) is given by,

kfriction (13Xi)

Nkatt (Zd -Zcm)

i=l

(For the simplicity, we use zi and Zi instead of aXi and /3Xi

respectively.)
F fRriction is the resultant friction force on the whole swarm,
given by;
N
onF
Nkf riction (im)
Fr friction
icti~~ricinT)

i=l

The term F/RboL corresponds to the resultant individual forces

acting on each robots, and it is given by,
N

zj)
.zjll3

(zi

0
j=lj7Xi
Therefore the final expression for the Resultant force acting
on the swarm is reduced to,
Frobot

FR

1J 1

iZ

i=l

Nkatt (Zd -Zcm) -Nkfriction (zcm)

which consists of two artificial force components, first

one being the active component which exerts a force on
Zcm towards Zd and the second one being the inactive
component depends on the velocity of Zcm and behave as the
resistive force. In above expressions, Zcm is the center of
mass of the swarm and 4cm is velocity of the center of mass
N

Z (zi)

krobot C R+ is a constant determining, the power of inter

individual repulsion and the distance between each other at
stable state. Bounds for krobot is discussed in section IV, with
respect to the stability of the formation.

Ffriction

friction

FR = Ftt

(4)

kmass C R+, represents the mass of the robot, and considered

to be a constant for all the members in the MAS.
The term Fi is the artificial force acting on each individual,
which makes them move toward the desired location (Zd) and
settles around there. Friobot term in the control function avoids
the inter-agent collisions and this function looks similar to the
repulsive term described in [8], [13] etc, but the problem of decreasing repulsion behavior for infinitesimally small distances
is avoided, thus inter-agent collision avoidance is guaranteed.
Term FJrictionis the artificial friction force exterted on each
agent, this force term makes the agent to a complete stop when
the forces are balanced, or when stable state is reached.
Before analysing the behavior of the MAS, we define the
artificial forces as follows. (note that for notational simplicity,
we use Fatt, Fjriction and
terms instead of functionals

Fatt (Zi, Zd),

-FR

+ robot

Where, FR is the resultant attraction force on the whole

swarm and is given by;

where

ui

In this section the behavior of the multi-agent system

as a whole (whole swarm) is analysed and conditions for
convergence obtained. Please note that we use term swarm,
to represent MAS in the following text. In our analysis, whole
swarm is considered as one complete object, where motion is
governed by the resultant force, FR,

of swarm, and are defined as, Zcm

ZNcm

A. Stability and convergence

Let e to be the error between target position and the current
position of the swarm, such that;

(7)
Page 130

6=

(Zcm -Zd)

motion towards Zd
200

Proof:
Direction of motion of the swarm
If we select a Lyanonov function candidate as Vcm =
1
1
k 2, then
will be bounded by,
M l2

150

100
50

Vcm <

-C|e|2

2 > 0 and Vcm < 0, one can say that the

Thus as Vcm
2
motion of the swarm is in the direction of decreasing c.
e

kfriction

Fig. 1.

Conditions for stability at the destination

The model described in (8) can be converted to standered
expression for frequency domain analysis as,

0.2

1
20

G(s)

15

s2+

C10
.

(9)

Then using the general frequency domain second-order transfer

function,

-5

+ +2

-10
2

-15

nw

2(bwnS

-31

Fig. 2.

kfriction

1.5

10

15

30-

25
20
15

15

-5

x coordinate

Fig. 3.

then e
of the

Zcm and e

kfriction

Fig. 4.

Nkmass, K

= -C

e-Kc

Nkatt and C

10

-E02

(8)
=

1D

3o 0
a-10

Nkfriction.

-M

Theorem 1: The center of

the center of
become stable at Zd,
converge to

Zcm=

where M

30 r

Therefore the equation of motion

swarm, with respect to error, is as follows;
=

kfriction

mass

mass

of the swarm, (Zcm) will

of the contour, (Zd) and

kmass, katt, kfriction > 0, and

if kfriction < 20kmass katt, the swarm demonstrates
under damped dynamics around Zd and,
(b) if kfriction > 2V/kmass katt, it will demonstrate damped

Li

if

-20 _

-30 F

Simulation steps

(a)

Fig. 5. Behaviour of the error between current and disired locations of Zcm
for some kf rictionvalues

Page 131

we obtain the corresponding value for ( from (9) is as below,

kfriction

2 v M-K

2 /kmass katt

(10)

which proves our assertion.

a
Note that in proving above, we consider the vector based
system (8) to be a scalar second order ODE. This is possible
as all the artificial forces are acting along the vector e and we
can reduce the vector equation to a scalar system assuming a
linear motion in the direction of e.
Above Theorem says that the swarm described by (1) with the
control function described by (4) will move to the target point
Zd, but it does not say anything about the motion of individual
members.
IV. BEHAVIOR OF A SINGLE AGENT
In this section we investigate the behavior of an individual
member of a specific swarm, where members have approximately equal inter-member repulsion function magnitudes. We
call such a swarm, "X Swarm".
Definition 1: A swarm is defined as "X swarm", if there
exists positive constants A, A, 6, , that satisfy the following
conditions simultaneously for all i, j and i t j.
1) dij > +A,
2) For dij > 6, /3ij , (Constant)
lZi- zcmil

603

4)
c
2heellzi - zcij
Where

(60
<A

dij

Proof:

Frobot

I{
Frobt =: krobot
krobotK
Frobot

1,ij

=
(1 1)
and Zi
Nii 1
In the above definition, Zim is the center of mass of the swarm
without the jth member. Also the condition l*z-Zi Tmj < A
sets a limit for an agents' position relative to other agents.

Remark 1: The repulsion force between individual agents

can be described as follows, the term -i
being the
zj~
directional component consists of a unit vector along the force;
1
and the term 1l
1 2 being the magnitude of the force
which is inversely proportional to the distance between them.
*

1) (zi

(N
Fi0b0L
Frobot < krobot 6G3

(z,Zi

C'M)

as (

Zcm),

'Zcm)

Z* Zicn~~~ Z cTJ

B. Motion of a member in "X swarm"

Using the dynamic model for the swarm, introduced in (1),
the motion of a member can be described by,

ii

katt

kmass

-zi) +

(Zd

i)3 -kfrictionriZ

/j~

(12)

We define the error between Zd and Zcm as, ei = Zi- Zd,

resulting ej = Zi and ej = Zi (as Zd remain constant). In
Theorem 1, we proved that the center of mass of the swarm
(Zcm) moves towards Zd and become stable at Zd. For our next
result, we use the fact that Zcm moved to Zd, i.e Zcm = ZdTheorem 2: Consider a "X Swarm", where Zcm = Zd at
any given time; if katt> (N 63 1), then the motion of each
krobot
member is towards the target location(Zd).
Proof: From (12) we can generate equation of motion
for a member in "X Swarm" with substituting ei terms as,

kI
kmass

katt ei +

krobot

zj)

Iz: ZJ V1
j=l,j7 i
E

kfrictionri) (13)

Choosing a Lyapunov function candidate for the member i as

A. Maximum Repulsion Force on a member in "X swarm"

< krobot(N -1) ((Z

6G3

* Zj)

, + A ) 3, and using (z, Zi- cm) < A (zi Zcm)

(6 + A)3'
~~~~~~~zi
-zcm

Taking

~zi

Vi= 2 kmasseiei +

Before investigating the motion behavior for the members

of "X Swarm", we introduce a bound for the artificial force
component Friobot in following Lemma.
Lemma 1: For a member in a "X Swarm",

J=1,J7 i

j=l1,ji

llZ-zj

(zi

jE

krobot ; (N -1) (zi- zm)

krobot

Fi0b0L
Frobot

(z3 z)

ij
J=l1,J7i
Using the definition of the "X Swarm", we have

A
+ A\)3

=krobot

2eiei (katt -krobot(N

1)

and using the following fact (from Lemma 1),

-C)
cm

Page 132

Zj)

(N

with Zcm is on Zd,

VI. CONCLUDING REMARKS

N

j=lj7i

(z,

zj)

~Zi

Zj

1)

(N
0

3<

1e 1

we can prove that Vi is bounded by,

Vi <- kfriction
Further

katt
krobot

> (N

-1)
d

Je

V >0.

Hence, the Lyapunov stability criteria is satisfied, which proves

the Theorem.
a
V. SIMULATION AND RESULTS

In this section, a computer simulation was used to illustrate

the results derived in previous sections and to demonstrate the
effectiveness and robustness of the swarm model for various
disturbances. We consider robots/members having point mass
dynamics with unit mass (i.e. k,,as = 1), throughout the
simulations.

A. Behavior of the whole swarm

First we illustrate the results derived in section III, i.e.
the behavior of whole swarm as a second order ODE. To
demonstrate the behavior with variation of kfriction (with
Zd = (-5, -3) and katt = 1), we present simulation results for a range of kfriction values (6, 2,1.5, 0.2) in figures 1 to 4 (representing different aggregation states; i.e
over/critically/under/zerol damped) and in Fig. 5 we demonstrate the variation of error ei against time.
B. MAS behavior in sudden failure/death of members
Swarms in real-life applications, especially in hostile environments, are subjected to face sudden erase of whole section
of the swarm from disasters, or failure/malfunction of some
members in random manner. In all those situations the existing
members should rearrange the formation in order to continue
the mission. The series of simulation screen shots presented
in Fig. 6. shows the behavior of proposed swarm model in
sudden loss of agents.

C. MAS behavior in recruiting new members to the society

Replacements of dead/malfunctioning members, increasing
number of members to stay with increasing demands etc,
results in adding new members to an existing MAS, which
may already in stable state. Thus the MAS should rearrange the
formation to recruit new members to the society. The series of
simulation screen shots presented in Fig. 7. shows the behavior
of proposed model in recruiting new members in to an existing
formation.
'In the zero damped situation, we include a small damping factor to
eliminate infinitely large movements of agents

In this paper, we present an aggregation algorithm, based

on artificial formation forces, for a multi-agent system or
swarm of robots with self localization and effective peer-topeer communication capabilities. We showed that the motion
of the whole swarm can be described as a second order ODE
and analysed the behavior of it with respect to stability and
convergence. Further, the motion and behavior of a member
was analysed with introducing a special swarm configuration
called "X Swarm".
Comparing with swarm aggregation models where members
having minimum sensing and computational capabilities (such
as proximity sensors, simple logic based controllers etc), our
swarm model needs more advanced sensing, communicating
and computing power. With recent advances in technologies,
such as wireless networking, GPS based self localization, etc.,
this system can be effectively used in swarm aggregation applications involving macro scale robots (e.g. land exploration,
land-mine removal etc,).
In our swarm model, the unbounded nature of the intermember repulsion force (F0bodt), at infinitesimally small intermember distances, may result implementation difficulties with
real-world actuators; but the same property, effectively avoids
inter-member collisions as it uses the maximum capacity of
actuators to repulse them selves away from each other.
REFERENCES
[1] C. Brown and K. Warburton, "Social mechanisms enhance escape
responses in schoals of rainbowfish," Environmental Biology of Fishes,
vol. Vol 56, pp. 455-459, 1999.
[2] I. Hamilton and L. Dill, "Group foraging by a kleptoparalastic fish: A
strong inference test of social foraging models," Ecology, vol. Vol 84(12),
pp. 3349-3359, 2003.
[3] D. M. Gordon, "The organization of work in social insect colonies,"
Complexity, vol. Vol 8(1), pp. 43-46, 2003.
[4] E. Robinson, D. Jackson, M. Holcomber, and F. Ratnieks, "'no entry'
signal in ant foraging," Nature, vol. Vol 438, p. 442, November 2005.
[5] D. Grunbaum, S. Viscido, and J. Parrish, "Extracting interactive control
algorithms from group dynamics of schooling fsh."
[6] Y. Inada, "Steering mechanisms in fish schools," Complexity International, vol. Vol 8, pp. 43-46, 2001.
[7] V. Gazi and K. Passino, "A class of attraction/repulsion functions for
stable swarm aggregations," Decision and Control, 2002, Proceedings
of the 41st IEEE Conference on, vol. Vol 3, pp. 2842-2847, 2002.
, "Stability analysis of swarms," Automatic Control, IEEE Transac[8]
tions on, vol. Vol 48, pp. 692-697, 2003.
, "Stability analysis of social foraging swarms," Systems, Man and
[9]
Cybernetics, Part B, IEEE Transactions on, vol. Vol 34, pp. 539-557,
2004.
[10] T. Balch and R. Arkin, "Behavior-based formation control for multirobot
teams," Robotics and Automation, IEEE Transactions on, vol. Vol 14,
pp. 926-939, 1998.
[11] 0. Soysal and E. Sahin, "Probabilistic aggregation strategies in swarm
robotic systems," Swarm Intelligence Symposium, Proceedings 2005
IEEE, pp. 325-332, 2005.
[12] K. Naruse, H. Yokoi, M. Kinoshita, and Y. Kakazu, "Group formation of
agents with two-dimensional inner state and one-to-one subjective evaluation," Proceedings IEEE International Symposium on Computational
Intelligence in Robotics and Automation ,, vol. Vol 3, pp. 1492- 1497,
2003.
[13] V. Gazi, "Swarm aggregations using artificial potentials and slidingmode control," Robotics, IEEE Transactions on [see also Robotics and
Automation, IEEE Transactions on], vol. Vol 21, Issue 6, pp. 1208-1214,
2005.

Page 133

-4
-6
-6 I_

.*

-8

.-2 -10

.-

-12

**

-14 1

-12

-14

-16
-220

* .

10

-16

-18

-16

-14

-12

-10
-8
x coordinate

-6

-4

-2

14

-4

-10
-8
coordinate

-6

-4

-2

-6

-4

C -2

-4

0 -2

10
x coordinate

(b)

(a)
1

12

* 0
-2

-12

-14-

-16

-16

-20

-18

-16

-14

-1
x

-10
-8
coordinate

-6

-4

-2

-220

-18

-16

-14

-12
x

(d)

(C)
Behavior of the MAS in loss of agents

Fig. 6.

-8

0*
0

z5

*
.
)
_
. * *

-10

*
*
*
>~~~ _0

-12-

-12

-14-

-14

-16-

-16

-20

-18

-16

-14

-1

-10

x coordinate

-8

-6

-4

-2

.*0

1.
7 -10

B1

20

-15

-14

-12

-10

x ord in at

(a)

(b)
-2

-8

i0_.
>-2

*S

.g
12

-12

-14

-14

-16

-16

-20

-18

-16

-14

-1

-10

x coordinate

-8

*._*X

-10

1*

-6

-4

-2

-220

.0

[_
-8

-16

~*
-14

-12

-10

x coordinate

(C)

(d)
Fig. 7.

Behavior of the MAS to adapt

Page 134

new

members

-8

B6