MOBILITY AWARE ROUTING WITH PARTIAL ROUTE

PRESERVATION IN WIRELESS SENSOR NETWORKS

Sanat Sarangi, Subrat Kar
Bharti School of Telecommunication Technology
And Management, IIT Delhi
New Delhi 110016 India
sanat.sarangi@gmail.com, subrat@ee.iitd.ac.in


ABSTRACT
We propose a novel reactive routing algorithm called PARTROUTE which achieves
energy efficient routing in sensor networks with stationary and mobile nodes.
PARTROUTE uses coarse mobility awareness in each node to reduce the overheads of
reactive routing by intelligently re-using parts of the created on-demand routes beyond
their standard lifetime. We show that the responsiveness of the network improves while
the overall network traffic is reduced. The effectiveness of PARTROUTE with a large
percentage of mobile nodes is also shown.


Keywords: wireless sensor network, routing protocol, PARTROUTE, mobility
aware, AODV, AODVjr


1 INTRODUCTION

A typical sensor network has a single gateway
and a set of connected nodes that participate in
sensing and forwarding application defined events to
the gateway. Some of the key driving factors in
designing routing algorithms for sensor networks are
(a) size of the network (the number of nodes), (b)
limited node energy, (c) deployment topology and
(d) provision for node mobility. Traditionally, there
have been two approaches to routing proactive and
reactive. In proactive routing, routes can be
computed and updated beforehand so that data is sent
through the best possible route. In reactive routing,
routes are created on-demand. The diversity of
sensor network applications have driven the
evolution of classes of application appropriate
routing algorithms which can be cluster-based [1],
hierarchical [2], robust [3], energy-aware [4],
geographic [5], query based [6] or swarm-behaviour-
based [7].
When some nodes in the network are mobile,
reactive routing becomes necessary. Dynamic Source
Routing (DSR) [8] and Ad-hoc On-demand Distance
Vector Routing (AODV) [9] are popular reactive
algorithms developed for mobile ad-hoc networks
(MANETs).

2 MOTIVATION

Most reactive algorithms assume that the nodes
in the network are mobile. However, mobile nodes
are not explicitly aware of their own mobility.
Recently, in [10], PH-MA-AODV and Agg-AODV
have been proposed that use mobility awareness to
improve the performance of reactive algorithms with
high-speed MANETs. In PH-MA-AODV a highly
mobile node discards a route-request to prevent itself
from participating in route formation. In Agg-AODV
the route-request accumulates the degree of mobility
of each intermediate node and the destination
chooses the route with minimum mobility. However,
both the algorithms assume that the intermediate
nodes can precisely measure their speed of
movement. In [11], prediction of broken links is used
to trigger routing updates before links break.
Mobility requirements in most sensor networks
are different from those of MANETs. The emphasis
in sensor networks is on minimizing energy
consumption with provision for (often low) mobility
within a stationary mesh of nodes where all nodes
talk to a central gateway. We propose a routing
algorithm PARTROUTE for memory-constrained
and energy-sensitive sensor networks with stationary
and mobile nodes where each node is explicitly
aware of its mobility status at any point of time and
this knowledge is used to intelligently preserve
part(s) of the route(s) formed during reactive routing
beyond the standard route lifetime. PARTROUTE
dynamically adapts itself to changing network
conditions and helps improve event delivery while
reducing both (a) the overall network traffic and (b)
latency of route formation (which is critical) for
mobile nodes.
With rapid strides in Micro Electro-Mechanical
Systems (MEMS), coarse mobility (moving or
stationary) awareness can now be easily added to

sensor nodes with the help of low-cost
accelerometers. Accelerometers have been used to
detect gait postures like walking, standing, lying [12]
[13]. Such information in the context of the
behavioural pattern of the subject can be used to
know whether a node is (and to predict if is likely to
remain) stationary or mobile. In a sensor network
containing such semi-mobile nodes, any semi-mobile
node that finds itself stationary could temporarily
behave like any other (always) stationary node.
PARTROUTE is designed to effectively use this
behavior to significantly improve routing
performance in such networks.
The rest of the paper is divided into four sections.
Section III describes the algorithm. In Section IV we
evaluate the performance of PARTROUTE
analytically and compare it with the simulation
results given in Section V. PARTROUTE is also
compared with AODVjr [14] to show the improved
performance under different conditions.

3 THE PARTROUTE ROUTING
ALGORITHM

In PARTROUTE, we consider a sensor network
with N nodes v
1
..v
N
having a set of stationary nodes
V
S
and a set of mobile nodes V
M
. |V
S
| + |V
M
| = N and
V
S
r

V
M
= N. Any node v
i
belongs to only one of the
sets and adheres to the policies laid out for that set.
L(v
i,
v
j
)

is a link between v
i
and v
j
, i = j. In
PARTROUTE we have two kinds of links static
and dynamic. If two nodes v
i,
v
j
e V
S
, they are
permitted to form a static or a dynamic link. In all
other cases the link is dynamic. Links are created and
updated only on reception of packets and a table is
used to maintain link information. A route between
v
i
and v
j
with intermediate node v
k
is denoted by
P(v
i,
v
k,
v
j
), i=j=k. A dynamic link has a lifetime of
lifetime
DYNAMIC
and for a static link it is lifetime
STATIC

with lifetime
DYNAMIC <
lifetime
STATIC
. The value for
lifetime
STATIC
is application specific and is further
discussed in Sections 4 and 5. It depends on factors
like deployment area, number of nodes, mobility
behaviour of nodes and their event generation rates.
AODV is an established reactive routing
algorithm for MANETs and AODVjr is a successor
to AODV which achieves performance comparable
to AODV without overheads such as sequence
numbers, gratuitous RREP, hop count, hello
messages, RERR and precursor lists from AODV.
AODVjr is therefore more suitable for memory and
energy constrained sensor networks. We compare
PARTROUTE with AODVjr.
PARTROUTE uses two kinds of route-requests:
M-ROUTE-REQUEST and S-ROUTE-REQUEST
for route creation as discussed in Section 3.4 and
Section 3.5 respectively. M-ROUTE-REQUEST is
processed by v
i,
v
j
e V
S
UV
M
whereas S-ROUTE-
REQUEST is processed by v
i
if v
j
e V
S
and
discarded if v
j
e V
M
. A mobile node uses only M-
ROUTE-REQUEST to create a route to the gateway.
All stationary nodes in the network except the
gateway use S-ROUTE-REQUEST to reach the
gateway. The gateway uses M-ROUTE-REQUEST
to find a stationary or mobile node in the network
since M-ROUTE-REQUEST is accepted and
processed by all nodes. We assume that the network
has a single gateway and each node in the network is
aware of this gateway. The principle can be extended
to multiple gateways, without loss of generality.
In PARTROUTE stationary nodes are allowed to
behave like mobile nodes under certain conditions. If
the duration for which mobile nodes rest (i.e. become
stationary) is large when compared to the duration
for which they move, they can be allowed to
temporarily function as stationary nodes during the
stationary phase. The target application determines
(a) the use of this role-switching feature and (b) the
duration of rest. Role-switching is discussed in
Section 3.3.

3.1 Reactive Route Discovery
We use Expanding Ring Search based route
discovery [15], an energy efficient method to
discover a route to a destination node. The source
sends a route-request message with TTL set to 1 and
simultaneously starts a timer that expires after twice
the node-traversal time. If the destination is one hop
away it sends a route-reply message which reaches
the source before the timer expires. If a route-reply is
not received, the source increases the TTL value,
proportionately increases the timeout and repeats the
process till either the destination is reached or the
source gives up. This limits the propagation extent of
the flooded route-request thereby saving energy. The
destination sends the route-reply back to the source
along the route taken by the first route-request it
receives. When a node with a valid route to the
destination sends a route-reply on behalf of the
destination, the reply is called gratuitous [15].

3.2 Partial Route Preservation
Partial route preservation implies retaining a part
of a reactively created route, for a period longer than
the standard lifetime for the route. The partial route
is called a trace. Each node on a trace is called a
representative. A stationary or mobile node which is
not a representative is called a non-representative.
The gateway is considered to be a representative of
itself. In PARTROUTE there can be two kinds of
route-replies to a route-request: (a) intermediate-
node route-reply that can only be generated by a
representative, (b) destination-node route-reply that
only the gateway can generate. Henceforth, a node
that is allowed to send a route-reply for a route-
request is called a reply-node. It may be noted that an
intermediate-node route-reply is different from a
gratuitous route-reply [15] which is generated by any

node wit
sequence
the node
Sequence
PARTRO
gratuitou
any com
intermedi

Figure 1:
Fig. 1
of station
mobile n
which the
the route-
till it rea
becomes
candidate
create-tra
Statio
route afte
represent
the route
stationary
about it.
gateway
represent
than the
stationary
requests e
Henc
of rep
respectiv
non-repre
set of po
nodes in t

3.3 Cr
A ro
called rep
th a route t
number info
and that rec
e numbers
OUTE nor AO
s route-reply,
mmunication
iate-node rout
Trace formatio
1 illustrates tr
nary (black) an
node 6 sends
e gateway rep
-reply (RREP
aches a mob
a representa
es are conve
ace messages
onary nodes (n
er a mobile n
tatives. The m
e-reply with t
y nodes that p
Any subsequ
by a mobile
tatives which
e gateway cr
y nodes. Thi
even faster.
ceforth, we us
presentatives
ely. A repres
esentative by
ositive intege
the network).
reate-trace M
oute-reply cre
presentative c
to the destin
ormation alrea
ceived with t
are used
ODVjr. More
a representat
to the gate
te-reply that it

on in PARTROU
race formation
nd mobile (wh
s a route-req
plies. Each sta
P) encounters
bile node (no
tive candidat
erted to repr
discussed in S
node 5 in Fig.
node (node 4
mobile node (
this informati
process the ro
uent request f
node can be
are closer to
reating more
is helps servi
se S
R
and S
U

and no
sentative is de
U
i
, i e I
+
, whe
ers. S
R
U S
U
=
Messages
eates a set of
andidates whi
nation based
ady present w
the route-requ
d by nei
eover, unlike
ive does not s
eway about
t generates.
UTE
n in the prese
hite) nodes wh
quest (RREQ)
ationary node
on its return p
ode 4 in Fig
te. Representa
resentatives w
Section 3.3.
1) that occur
4) cannot beco
(node 4) upd
ion to notify
oute-reply afte
for a route to
e serviced by
the event sou
e traces on
ice future ro
to denote the
on-representat
enoted by R
i
ere I
+
denotes
= N (number
f potential no
ich could beco
on
with
uest.
ither
the
send
the

ence
here
) to
that
path
. 1)
ative
with
in a
ome
dates
the
er it
the
the
urce
the
oute-
e set
tives
and
s the
r of
odes
ome
rep
rep
nod
gat
bec
can
the
get
can
trac
the
cre
rou
the
is n
trac
ack
con
trac
upd
cre
retr
ack
Ack
to
inte
trac
bec
Tra
rep
rep
ene
this
that
it d
wit
0 <
ove
can
inte
gat

3.4
RE
RO
mo
If a
nod
gat
hav
cre
stat
swi
nod


presentative no
presentative ca
de 3 in Fig.
eway and rec
come a repr
ndidate does n
path of the ro
s the route-re
ndidates sends
ce-request (C
representativ
ate-trace. Wit
ute request ori
last represent
not used here.
ce acknowle
knowledgemen
A create-trac
nvert all dynam
ce originator
dating their l
ate-trace orig
ries to send c
knowledgemen
knowledgeme
(a) lossy lin
ermediate stat
ce (or its ack
comes mobile
ansmission o
presentative f
ply that it gen
ergy consump
s behaviour b
t every time a
does not send
thin the next
< CT
PERIOD
erheads due to
n also becom
ermediate rep
eway about th
Route Fo
Stationar
Stationary
EQUESTs to
OUTE-REQUE
de operation.
an S-ROUTE-
de, it is simply
eway with S-R
ve only statio
ated in the pr
tionary node f
itching (Secti
de (Section 3.5
odes (Section
andidates form
1) sends a cr
ceives an ackn
resentative. S
not know whet
oute-reply, th
eply after a s
s a special m
TR) to the pre
ve candidates
th no intermed
iginator sends
tative candida
. The gateway
dges its rec
nt to the creat
ce message an
mic links on th
and the gatew
ink lifetimes
ginator makes
create-trace if
nt befor
ents may not
nks or (b) ro
tionary node
knowledgeme
and voluntari
of a create-
for every int
nerates can sig
ption level of
by defined a v
a representativ
a create-trace
(lifetime
STATI
lifetime
S
o create-trace.
me carriers
presentative ca
heir presence.
ormation An
ry Nodes
nodes onl
create route
ESTs only su
So, gateway
-REQUEST i
y discarded. S
ROUTE-REQ
onary nodes i
rocess is the
fails to form a
on 3.6) and
5).
3.2). The las
med by a route
reate-trace (C
nowledgemen
Since a repr
ther it is the la
e first mobile
series of repr
message called
evious node (
s) asking it t
diate mobile n
s the create-tra
ate. Create-tra
y on receiving
ceipt by sen
e-trace origin
nd its acknowl
he path betwe
way into static
to lifetime
ST
s a suitable n
f it does not r
re giving
be received e
le-switching
on the path
ent) message
ily drops the m
trace messag
termediate-no
gnificantly in
the network.
variable CT
PE
ve sends a cre
e for any even
IC
- CT
PERIOD
)
TATIC
to re
. Create-trace
of the nod
andidates to
nd Route R
ly use S
es to the gat
upport destina
is the only re
is received by
So a route crea
QUEST is gua
implying that
same as the r
a route, it perf
behaves like
st of these
e-reply (i.e.
CT) to the
nt (CA) to
resentative
ast one on
node that
resentative
d a create-
(the last of
to send a
nodes, the
ace if it is
ace request
g a create-
nding an
ator.
ledgement
een create-
c links by
TATIC
. The
number of
receive an
g up.
either due
where an
of create-
suddenly
message.
ge by a
ode route-
crease the
We limit
ERIOD
such
eate-trace,
nt received
) seconds,
duce the
messages
de ids of
notify the
epair for
S-ROUTE-
teway. S-
ation-node
eply-node.
y a mobile
ated to the
aranteed to
the trace
route. If a
forms role
a mobile

3.5 Ro
M
Mobi
form rou
indicate e
mode of
The inter
destinatio
circumsta


Figure 2:
extends tra
In
represent
ROUTE-
successfu
and a rep
terminatin
links be
represent
Even du
message
gateway
traces ge
principle.
from nod
ROUTE-
2, 6, 10)
to extend
After
mode, if
receive a
an applic
ROUTE-
a certain
RETRY-
as determ
gateway
created b
possibly
applicatio
nodes sw
ROUTE-
Since
all nodes
oute Format
obile Nodes
ile nodes us
utes to the g
either interme
f operation in
rmediate-node
on-node mod
ances as discu
M-ROUTE-RE
ace L(1, 2) to P
the interm
tative can be
REPLY to a
ully created ro
presentative co
ng at the repre
yond it or
tative is the ga
uring trace
acknowledg
instead of th
t extended by
. In Fig. 2 an
de (representa
REQUEST fr
to the gatew
d the trace L(1
creating a r
the mobile n
an acknowled
cation level
REQUESTs i
n number of
COUNT) usin
mined by th
is the reply-n
between the m
creating anoth
on level comm
witches back to
REQUESTs f
e M-ROUTE-
s they may
tion and Ro
se M-ROUTE
gateway. A m
ediate-node or
n the M-ROU
e mode is use
de is used
ussed below.
EQUEST from m
P(1,2,6)
mediate-node
a reply-node
an M-ROUTE
oute between
ould extend t
esentative by
create a ne
ateway itself a
extension,
ement is ge
he intermediat
y the recursiv
M-ROUTE-R
ative) 2 in res
rom node 10 f
way and adds s
, 2) to P(1, 2,
oute in the i
node discover
dgement from
event, it sta
in destination
f retries (giv
ng the interme
he target ap
node for this
mobile node
her trace in th
munication res
o using interm
for creating su
REQUESTs
form shorter
oute Repair
E-REQUESTs
mobile node
r destination-n
UTE-REQUE
ed by default
under spe

mobile node 10
mode,
and send the
E-REQUEST
the mobile n
the previous tr
adding new st
ew trace if
as shown in Fi
the create-tr
enerated by
te node. Exis
vely applying
REPLY gener
sponse to the
forms a route
static link L(2
6).
intermediate-n
rs that it did
the gateway
arts sending
-node mode a
ven by EVEN
ediate-node m
pplication. S
mode, a rout
and the gatew
he process. O
sumes, the mo
mediate-mode
ubsequent rout
are processed
r traces than
for
s to
can
node
EST.
and
ecial
0
any
e M-
T. A
node
race
tatic
the
ig. 1.
race
the
sting
this
ated
M-
P(1,
2, 6)
node
not
y for
M-
after
NT-
mode
ince
te is
way
Once
obile
e M-
tes.
d by
n S-
RO
nod
How
pro
nod
3.6
the
tran
beh
stat
dep
ope
nei
eve
a m
idle
con
mo
by
beh
Fou
circ
wh
com
how
Wh
its
RE
nod
RO
RE

Fi

rep
R
Q,
pro
cha
form
stat
resp
mo
R
Q,
rou
pac
=1
app
by
U
2,
betw
OUTE-REQUE
des since they
wever, repres
ocess would h
des generating

Role Swit
A stationary
gateway th
nsmitting a S
have like a mo
tionary node
ployment or
eration possi
ghbouring sta
ents through V
mobile node c
e and the
nversion after
bile node as d
We now exp
a node partici
haviour of PA
ur cases are
cumstances e
en it creates
mbination of
w PARTRO
herever applic
role immediat
EQUEST. Rol
de from its
OUTE-REPLY
EQUEST.
igure 3: Refere
Case 1: L
presentative U
,
G). If U
k
sw
ocesses M-RO
ance to becom
med if K = P
tionary.
Case 2: If
ponse to the
bile then {M
,
G} L S
R
.
ute-replies to
ckets since it i
Case 3: If o
switches its
parent since M
R
1
which com
,
U
P,
R
i
} L
ween represen
ESTs. This is
y would anywa
sentatives on t
help provide fa
g M-ROUTE-R
tching
y node that is u
hrough other
S-ROUTE-RE
obile node (Se
es that hav
have beco
ibly due to
ationary node
V
S
U V
M
instea
ould become
target appl
r considering
discussed in S
plain using Fig
ipating in rou
ARTROUTE a
e used to
encountered b
s a route to
these cases m
OUTE handl
cable, the con
tely after proc
le-switching
responsibility
Y it receives
ence diagram fo
et us assum
U
k
lies on P(M
witches its rol
OUTE-REQU
me a represent
or all nodes
R
1
that sent
e M-ROUTE
M, U
1,
U
2, ,
U
R
1
cannot s
subsequent
s not a represe
one of the oth
role, its effe
M-ROUTE-R
mes before R
i
.
L S
U
. If R
i
ntatives R
i-1
a
s acceptable
ay change the
the trace crea
aster response
REQUESTs.
unable to find
stationary n
EQUEST is a
ection 3.5). Th
ve been iso
ome isolated
o non-availa
es to be able
ad of only V
S
.
stationary if
lication justi
g the dynami
ection 2.
g. 3 how role-
ute formation a
after the route
analyze the
by a mobile
o the gatew
may be used t
les other
ncerned node
cessing the M
does not ex
y to process
for this M
or role-switching
me a station
M, U
1,
U
2, ,
U
P
le immediate
UEST, it for
tative after th
between U
k
a
the intermed
E-REQUEST
U
P,
R
1
} L S
U
an
send intermed
M-ROUTE-R
entative any m
her representa
fect is not im
REQUESTs ar
However, no
, now a mo
and R
i+1
gener
to mobile
ir location.
ated in the
es to other
a route to
nodes by
allowed to
his allows
olated by
d during
ability of
e to route
. Similarly,
it remains
ifies this
cs of the
-switching
affects the
is created.
different
node M
ay G. A
to explain
situations.
e switches
M-ROUTE-
xempt the
s the M-
M-ROUTE-

g analysis
nary non-
P,
R
1,
R
2, ,
ly after it
regoes its
he route is
and R
1
are
diate-node
becomes
nd {R
2, ,
diate-node
REQUEST
more.
atives R
i
, i
mmediately
re fielded
w {M, U
1,
bile node
rates a M-

ROUTE-
node resp
gateway
between
nodes th
intermedi
for event
this as a
node M
REQUES
breaking
new route
in the cy
link lif
communi
There
could adv
So we co
mobile
condition
Case
becomes
ROUTE-
represent
to R
1
) or
Else, U
K


4 PER
PAR

A large d
reactive a
the earlie
successfu
In this Se
to charac
respect to
used to a
Section 5

4.1 As
The nod
transmiss
carried ou
by a mob
C
K
with r
position
route-req
negligible
condition
for every
negligible
at the me
(denoted
any nod
transmitte
and proce
since they
REQUEST, R
ponse without
passes throu
R
i
and R
i-1
.
hat use rout
iate link fail
ts they sends
broken link
M-ROUTE-RE
STs) to find
the above cy
es do not use t
ycle are not us
fetimes exp
ication throug
efore, rapid ro
versely affect
onstrain fidget
to minimize
ns.
4: If a mobile
stationary j
REQUEST
tative provide
all nodes betw
remains a stat
RFORMANCE
RTROUTE
delay in route
algorithms due
er sections, w
ully help reduc
ection we buil
cterize PART
o delay in ro
analyze the s
5.
ssumptions
des are iden
sion range o
ut with respec
bile node M m
radius K as s
of M in the
quest and ge
e. All transm
ns are ideal so
y route-reques
e when comp
essage source
by , >0) fo
de within a
er is essentia
essing delays
y are applicab
R
i-1
could giv
t realizing tha
ugh R
i
, thus
When mobil
es with L
to receive ac
to the gatewa
problem and
EQUESTs
new routes
ycle in the pr
the nodes in th
sed, the cycle
ire due t
gh the links.
ole-switching b
performance
ty mobile nod
e the overh
e node U
K
swi
ust after pro
it either
d K = P (sinc
ween U
K
and
tionary non-re
E EVALU
e formation is
e to their on-d
we have seen
ce this delay f
ld an analytic
TROUTEs p
oute formatio
imulation res
ntical with a
of radius L.
ct events sent
moving within
shown in Fig.
e interval bet
etting back
mission, chann
o a route-resp
st. Radio prop
pared to the tr
e. So, the Hop
for a message
radius L o
ally the same.
are considered
ble to every no
ve a intermedi
at its route to
creating a cy
le (or station
(R
i-1
,

R
i
) as
cknowledgem
ay, they perc
d use destinat
(or S-ROU
to the gatew
rocess unless
he cycle. If no
e disappears a
o absence
by a mobile n
of PARTROU
des to always
heads due s
itches its role
ocessing the
becomes
ce U
P
is adja
R
1
are station
epresentative.
UATION
s a bottleneck
demand nature
n how traces
for mobile no
al model in o
performance w
on. The mode
sults presented
an ideal circ
The analysis
to the gatewa
n a circular reg
4. The chang
tween sendin
its response
nel and recep
ponse is recei
pagation dela
ransmission d
p Traversal T
to be received
of the mess
. Mean queue
d to be a part
ode. We assum
iate-
o the
ycle
nary)
an
ments
eive
tion-
UTE-
way
the
odes
after
of
node
UTE.
stay
such
and
M-
a
cent
nary.
OF
k for
e. In
can
odes.
rder
with
el is
d in
cular
s is
ay G
gion
ge in
ng a
e is
ption
ived
ay is
elay
Time
d by
sage
eing
of
me a
den
suc
gat
nod

Fig
ana
4.2
is a
rou
exp
M a
wh
wh
del
at d
B
x
.
Res
any
cum
x

RT

req
nod
nod
pro
the
leas
cas
wou
of t
by
rep
nse uniform d
ch that an eve
eway in x/L
des on a linear
ure 4: A circul
alysis
Estimatin
Time
Suppose at a
at a distance
ute to the gatew
pects a route-r
at a distance x
ich it originat
ich it receive
ay experience
distance x and
2xux
.
Hence
sponse Time
ywhere in
mulative delay
K in C
K
. Thi
T
A =
] D
x
.2
K
0
] 2x
K
0
With the ass
quest from M
des within a c
des again broa
ocess. In AOD
route request
st |xL] hops
se situation fo
uld have to fi
traces. For thi
For PARTRO
an intermedi
presentative co
deployment o
ent transmitte
hops through
r path between

lar sensor netwo
ng Route F
a given instan
x from the G
way. It genera
response. Dela
x is the durati
tes the route-r
s the route-re
ed by M for al
d (infinitesim
e, for a given
(RT
A
) experie
C
K
is obtain
y over all posi
is is given as
xdx
xdx

sumptions in
is received a
circle of radiu
adcast the rou
DVjr only G is
t. So, the route
to reach G.
or PARTROU
ield all route-
s case B
x
= 2
OUTE, the ro
ate node if it
ould be encou
of sensor nod
ed by M can
h (x/L1) int
n M and G.
ork deployment
Formation
nt T, the mobil
G and does n
ates a route-re
ay (B
x
) exper
ion between t
request and th
eply. So the c
ll its positions
mal) width dx
algorithm A,
enced by M p
ned by avera
itions at distan

Section 4.1,
at the same ti
us L around
ute-request to
s allowed to r
e request wou
This is also
UTE since the
requests in th
|xL].
ute-reply may
t is a represen
untered at n ho
des in C
K

reach the
termediate

t for
Response
le node M
not have a
equest and
rienced by
the time at
he time at
cumulative
s on a ring
from G is
the mean
positioned
aging the
nce x, 0
(1)
the route
ime by all
M. These
repeat the
respond to
uld need at
the worst
e gateway
he absence
y be given
ntative. A
ops where

1 n |x
lifetimes
formation
attempts
Assuming
equal pro
B
x
= (2
In the
the first h
Lowe
possibilit
choked w
performa
Response
speed mo
5.2.2). W
(TI) whic
PARTRO
the valu
PARTRO

TI =
RT
P
RT

4.3 Ef
The p
the numb
traces ar
preserved
faster res
thereby c
packets. I
gateway,
lifetime
refreshed
lifetime
ST
Deplo
area and
used to e
order to h

5 SIMU

Discrete
1000 seco
PARTRO
on PART
for PAR
Finally, T
a measur
PARTRO

5.1 Co
An i
nodes. T
always a
transmits
gateway.
xL]. Mobility
are config
n depends
and percen
g n takes eac
obability, the

n=|xL
n=1
n)
]
|
e best case, a
hop. So B
x
= 2
er the Resp
ty of the gate
with route-requ
ance in a live
e Time impr
obile nodes
We also define
ch is used to
OUTE when c
ue of TI, b
OUTE.
PARTROUTE
T
AODV]

ffect of Multip
performance
ber of event so
re created a
d for longer du
sponses to rou
controlling th
If a node in t
for this tra
of the netwo
d with event
TATIC
.
oyment param
the mean rate
stimate a suit
have effective
ULATION
event simula
onds is used t
OUTE. The ef
TROUTE is a
TROUTE are
Trace Index di
re of the Resp
OUTE.
onditions
ideal circular
There are no
awake. For
, it receives a
Non-receipt
y pattern of M
guration spec
on earlier
ntage of st
ch value with
mean value o
|xL] = (|x
all route-reque
2.
ponse Time,
way and nod
uest packets th
e scenario. A
roves perform
(further discu
e a figure of m
compare the
compared to A
better the re

ple Event Sou
of PARTROU
ources increa
and (b) exis
urations. Mor
ute-requests fr
he flooding o
the network h
ace to persis
ork it is imp
ts at interva
meters especi
e of event gen
table value for
traces.
ation with sim
to evaluate th
ffect of multip
analyzed. The
e compared a
iscussed in Se
ponse Time o
radio model
packet losse
every event
an acknowled
of acknowled
M is random, tr
cific and tr
route forma
tationary no
hin its range w
of D
x
is given
L] +1).
ests are fielde
lesser is
es near it get
hat could degr
Also, decrease
mance for h
ussed in Sec
merit Trace In
e effectivenes
AODVjr. Sma
esponsiveness

urces
UTE improve
se since (a) m
sting traces
e traces help g
rom mobile no
f routing con
has a trace to
st for the en
portant that i
als smaller t
ially size of
neration should
r lifetime
STATI
mulation runs
he performanc
ple event sou
e results obtai
against AOD
ection 4 is use
f the network
l is used for
es and nodes
t that the n
dgement from
dgement lead
race
race
ation
odes.
with
n by
ed at
the
tting
rade
e in
high
ction
ndex
s of
aller
of
(2)
es as
more
are
give
odes
ntrol
o the
ntire
it is
than
the
d be
IC
in
for
ce of
urces
ined
DVjr.
ed as
k for
r all
are
node
m the
ds to
retr
eve
dep
be
mo
cha
Mo
No
betw
loc
and
stat
stat
nod
the

Fig
dep
Q
K
(Fig
eve
eac
gen
gat
eac
we
incr
dist
hom
netw
sho
hol
(cen
also
PA
the
the
con
sec
dep
ransmission o
ent. A single
ployed non-ev
static or mob
re event-gen
aracterize the
ovement of m
des randomly
ween 1.0m/s
ation with th
d velocity are
tionary node
tionary node
de (discussed
duration of si
ure 5: A quarte
ployed nodes an
Nodes are d
(a quarter of
g. 5). Q
K
is d
ent-generating
ch node (L) i
nerated by a
eway needs 3
ch other by 75
use 10 node
rease the d
tribution of
mogeneous ev
work of reas
own that Trac
lds for each o
ntered at G w
o serves a
ARTROUTEs
gateway is a
network.
Each non
nfigured to be
ctor. Event-ge
ployment area
of the same ev
gateway, a s
vent- generatin
bile as per re
erating mobi
e performanc
obile nodes is
y choose a lo
s to 2.0m/s.
he chosen velo
e chosen. It
e may be e
or (b) a role-
in Section 2)
imulation.

er-circular regio
nd a gateway
eployed in a
C
K
in Section
divided into 3
g nodes in eac
s 75m. From
node farthest
3 intermediate
5m to reach t
es in each se
density of n
nodes across
ven with mob
sonable size f
e Index for C
of its four com
with radius K)
as a refer
performance
at a corner ra
n-event-genera
e mobile mov
nerating node
a. The range o
vent instead o
set (R) of 30
ng nodes (con
equirement) an
ile nodes are
ce of PART
s random way
ocation and a
After movin
ocity the nex
may be not
either (a) an
-switching sem
which is stat
on with random
quarter-circu
n 4.1) with rad
3 sectors with
ch sector. The
m Section 4.1,
t (i.e. 300m)
e nodes separ
the gateway.
ector to (a) su
nodes, (b)
s the deploym
bility and (c)
for analysis.
C
K
(in Section
mponent quar
) and hence fo
rence for
e in topologi
ather than the
ating nodes
ves only withi
e(s) move in
of each node
of the next
randomly
nfigured to
nd one or
e used to
TROUTE.
ypoint [8].
a velocity
ng to the
xt location
ted that a
n always
mi-mobile
tionary for

mly
ular region
dius 300m
h 10 non-
e range of
, an event
from the
rated from
However,
ufficiently
keep the
ment area
) create a
It can be
n 4.2) also
rter circles
for Q
K
. Q
K

analyzing
ies where
centre of
s when
in its own
the whole
is kept at

75m to h
range of
lifetime
DY
and EVE
estimated
event tr
generatin
node in P
the retrie
uses d
(Section
Hencefor
PARTRO
We u
the total n
transmitte
to denote
by all no
events se
an acknow
transmitte
transferre

TPE =
To

5.2 Re
5.2.1 E
From
reduce th
discussio
of event s
periodica
lifetime
ST
performa
significan
by increa
Figure 6:
lifetime
STA
200 secon
5.2.2 C
Resp
mobile ev
150s as
used for
have sufficient
f nodes. Pa
YNAMIC
is kept
ENT-RETRY
d Hop Traver
ansmission i
ng mobile nod
PARTROUTE
es defined by
destination-no
3.5) till it su
rth we use
OUTE with the
use the term
number of (da
ed by all node
e the total num
odes. Event C
ent by an even
wledgement.
ed for every
ed to the gatew
otaI Packct Co
Evcnt Count
esults
Effect of multi
m Fig. 6 it is cl
he Response T
n in Section
sources once t
ally, any furt
TATIC
does
ance e.g. for
nt improveme
asing lifetime
S
Variation of Re
ATIC
(lifetime
ST
ds)
Comparison of
ponse Time fo
vent source do
shown in Fig
r comparison
t overlap betw
acket size is
t at 15s, CT
PE
Y-COUNT is
rsal Time is
is attempted
de every 6 sec
E fails to deliv
y EVENT-RE
ode M-RO
uccessfully d
PARTROUT
e value of life
Total Packet
ata and routing
es. Event Pac
mber event pa
Count refers t
nt source for
So, the total n
y event (TP
way is given b
ount
t

iple event sou
lear that multi
Time of the ne
4.3. Also, for
the created tra
ther increase
not signifi
r a single e
nt in response
STATIC
from 150
esponse Time (
TATIC
takes valu
f PARTROUT
or PARTROU
oes not vary s
g. 6 So, PAR
n with AOD
ween transmis
s kept const
ERIOD
is set to
set to 3.
0.1 seconds.
by an ev
conds. If a mo
ver an event a
TRY-COUNT
OUTE-REQUE
delivers an ev
E-x to refer
etime
STATIC
as
Count to den
g control) pac
ket Count is u
ackets transmi
to the numbe
which it rece
number of pac
PE) successf
by

urces
iple event sou
etwork as per
r a given num
aces get refres
in the value
icantly impr
event source
e time is obser
0 to 200.
(in seconds) wit
es 50, 100, 150
TE with AODV
UTE for a sin
significantly a
RTROUTE-15
DVjr. A cer
sion
tant,
30s
The
An
vent-
obile
after
T, it
EST
vent.
r to
s x.
note
ckets
used
itted
er of
eives
ckets
fully
(S)
urces
r the
mber
shed
e of
rove
no
rved

th
0,
Vjr
ngle
after
0 is
rtain
per
mo
com
Fig
Tot
PA
nod
than
is
Res
eve
wh
two

Fig
AO
mob
are
the
inst
typ
for
(Fig
hig
Sec
mo
PA
del
to i
Fig
rcentage of th
bile and the p
mpared by va
gs 7, 8 and 9 s
tal Packet
ARTROUTE-1
It is observe
des, Response
n half of that
almost twice
sponse Time
en when up to
en almost all
o algorithm
ure 7: Variatio
ODVjr and PAR
bile nodes (0%
For 0% mo
sent over (th
stationary
tantaneously
ically shorter
PARTROUT
g 8) are due
gher Event Pac
ction 4.2, whe
ves at a (rel
ARTROUTE-1
ivering 3% m
its significantl

ure 8: TPE for
he stationary n
performance o
arying this per
show variation
Count and
50 and AODV
ed from Fig.
e Time for PA
t of AODVjr
e as respons
remains signi
40% nodes ar
nodes (80%
ms have s
on of Response
RTROUTE-150
: all nodes stati
bility, events
he long) trace
nodes where
created route
r. So, margin
TE-150 when
e to PARTRO
cket Count. H
en the event-g
latively highe
50 outperf
more events wi
ly lower Resp
r AODVjr and P
nodes from R
of both the alg
rcentage from
n of the Respo
d Event C
Vjr for 0-80%
. 7 that for
ARTROUTE-1
i.e. PARTRO
sive as AOD
ificantly smal
re mobile. As
nodes) are m
similar per
Time (in secon
for different nu
onary)
s in PARTRO
es created a p
eas in AOD
es to the gat
ally higher T
compared to
OUTE-150s
However, as di
generating mo
er) speed of
forms AOD
ith 21% lower
onse Time.
PARTROUTE-
are made
gorithms is
m 0 to 80.
onse Time,
Count for
% mobility.
stationary
150 is less
OUTE-150
DVjr. The
ller (60%)
expected,
mobile, the
rformance.
nds) for
umbers of
OUTE-150
priori over
DVjr, the
teway are
TPE levels
o AODVjr
relatively
scussed in
obile node
f 4-5 m/s,
DVjr by
r TPE due

150


With
decreases
PARTRO
improvem
lack of s
150 and
Count for
for all ca
almost sa
more eve
To se
with a lar
the simul
switching
The dura
given by
mean val
a standar
switching
is initiall
more eve
lower TP

Figure 9:
150
A to
deployme
behaviou
commonl
10 where
nodes an
through w
remaining
random-w
There i
Remainin
Transmis
between
PARTRO
compared
and 20%
node tran
in carryin
mobility, len
s. For 40
OUTE-150 i
ment of 26%
sufficient stat
AODVjr hav
r PARTROU
ases except w
ame. PARTRO
ents than AOD
ee the effects
rge stationary
lation we con
g its status be
ation for bein
y two normal
lues of 450s a
rd deviation
g from happen
ly mobile. PA
ents, has a 17%
PE compared t
Event Count fo
opology in
ent has been u
ur of PARTR
ly used confi
e 50% of the
nd the gatew
which events
g 50% move
waypoint mob
s one even
ng conditions
ssion range
adjacent node
OUTE-150 de
d to AODVjr
lower TPE. W
nsmission rang
ng out localiza
ngths of traces
0% mobili
is 74% of
. With 80%
tionary nodes
ve similar TP
TE-150 is mo
with 0% mob
OUTE-150 de
DVjr with 40%
s of semi-mob
y period withi
nfigure each n
etween mobil
ng mobile an
lly distribute
and 50s respec
of 10s. Thi
ning synchron
ARTROUTE-
% lower respo
o AODVjr.
or AODVjr and
Fig. 5 with
used to charac
ROUTE. How
guration wou
stationary no
way) form a
can be easil
throughout th
bility model d
nt-generating
s are the s
of nodes an
es on a row o
livers 12% m
at a 20% low
With appropri
ge, the configu
ation [16] [17]
s shorten so T
ity, TPE
f AODVjr
mobility, due
s PARTROU
PE levels. Ev
ore than AOD
bility where i
elivers up to 1
% mobility.
bility (Section
in the duration
node in R to k
le and station
nd stationary
d variables w
ctively, each w
is prevents r
nously. Each n
-150 delivers
onse time and
d PARTROUTE
h random n
cterize the gen
wever, a typ
uld look like
odes in R (i.e
connected
y routed and
he grid using
discussed ear
mobile no
same as ear
nd the dista
r column is 7
more events w
wer response t
iate adjustmen
uration also h
].
TPE
for
an
e to
UTE-
vent
DVjr
it is
15%
n 2)
n of
keep
nary.
are
with
with
role-
node
5%
d 8%

E-
node
neral
pical
Fig.
. 15
grid
the
g the
rlier.
ode.
rlier.
ance
75m.
when
time
nt of
helps

Fig
5
Ind
0.3
Fig
Ind
the
low
stat
Tra
0.6

Fig

6

We
into
(tra
incr
mo
imp
Per
con
app
cha
netw
dep
PA
ure 10: A grid
5.2.3 Variation
From Sectio
dex with L=75
2 for worst,
g.11 shows ho
dex for PART
oretical wors
west value of 0
tionary nodes
ace Index is 0.
6.
ure 11: Trace I
CONCLUSI
e have shown
o stationary a
ace) preservat
rease the ef
bility aware
portance of
rformance o
nditions has
plicability to
aracteristics o
work, nature
ployment topo
ARTROUTE

deployment of
n in Trace Ind
on 4.2, theor
5 and K=300 a
mean and b
ow the simul
TROUTE-150
st, mean and
0.44 is observ
s. With 40%
.6 which is be
Index Analysis
ION
how logicall
and mobile no
tion over stat
ffectiveness o
e sensor n
energy-effic
f PARTROU
been show
o different
of the nodes
e of event
ologies could
to bring
f nodes
dex
retical values
are given as 1
best case res
lation results
0 compare ag
best case le
ed for Trace I
mobility, the
etter than mea
ly dividing th
odes with par
tionary nodes
of reactive r
networks wh
ciency is p
UTE under
wn to illus
situations.
comprising t
traffic and
d be used to
about s

of Trace
, 0.66 and
spectively.
for Trace
gainst the
evels. The
Index with
observed
an value of

e network
rtial route
s can help
routing in
here the
aramount.
different
strate its
Mobility
the sensor
d typical
configure
significant


improvements on all routing performance parameters.
PARTROUTE can also find significant application
with Body Sensor Networks where the mobility
pattern of nodes can be determined with sensors like
accelerometers.

7 REFERENCES
[1] W. R. Heinzelman, A. Ch, and H. Balakrishnan:
Energy-efficient communication protocol for
wireless microsensor networks, IEEE Proc.
Hawaii Intl. Conf. Sys. Sci., pp. 110 (Jan.
2000).
[2] A. Manjeshwar and D. P. Agrawal: Teen: A
routing protocol for enhanced efficiency in
wireless sensor networks, Intl. Parallel and
Distributed Processing Symp., vol. 3, pp.
20092015 ( 2001).
[3] D. Ganesan, R. Govindan, S. Shenker, and
D. Estrin: Highly-resilient, energy-efficient
multipath routing in wireless sensor networks,
SIGMOBILE Mob. Comput. Commun. Rev.,
vol. 5, no. 4, pp. 1125 (2001).
[4] R. C. Shah and J. M. Rabaey: Energy aware
routing for low energy ad hoc sensor networks,
Proc. of the IEEE Wireless Communications
and Networking Conf. (WCNC), Orlando,
Florida ( 2002).
[5] B. Karp and H. T. Kung: Gpsr: greedy perimeter
stateless routing for wireless networks,
MobiCom 00: Proc. of the 6th Annual Intl.
Conf. on Mobile Computing and Networking,
New York, NY, USA: ACM, pp. 243254
(2000).
[6] C. Intanagonwiwat, R. Govindan, and D. Estrin:
Directed diffusion: A scalable and robust
communication paradigm for sensor networks,
in MOBICOM, ACM, pp. 5667 (2000).
[7] L. Wang, R. Zhang, and S. Geng: An energy-
balanced ant-based routing protocol for
wireless sensor networks, WiCOM09: Proc. of
the 5th Intl. Conf. on Wireless communications,
Networking and Mobile Computing,
Piscataway, NJ, USA: IEEE Press, pp. 3556
3559 (2009).
[8] D. B. Johnson and D. A. Maltz: Dynamic source
routing in ad hoc wireless networks, Mobile
Computing, ser., The Kluwer International
Series in Engineering and Computer Science,
T. Imielinski and H. F. Korth, Eds, Springer
US, vol. 353, pp. 153181 (1996).
[9] C. Perkins and E. Royer: Ad-hoc on-demand
distance vector routing, Proc. of Second IEEE
Workshop on Mobile Computing Systems and
Applications, 1999. WMCSA 99., pp. 90100
(Feb. 1999).
[10] Y. Khamayseh, O. M. Darwish, and S. A.
Wedian: Ma-aodv: Mobility aware routing
protocols for mobile ad hoc networks,
ICSNC 09: Proc. of the 2009 Fourth Intl. Conf.
on Systems and Networks Communications,
Washington, DC, USA: IEEE Computer
Society, pp. 2529 (2009).
[11] S. Ju and J. Evans: Mobility-aware routing
protocol for mobile ad-hoc networks, IEEE Intl.
Conf. on Communications Workshop, 2009, pp.
16 (Jun. 2009).
[12] M. Mathie, B. Celler, N. Lovell, and A. Coster:
Classification of basic daily movements using a
triaxial accelerometer, Medical and Biological
Engineering and Computing, vol. 42, pp. 679
687, 2004.
[13] Q. Li, J. Stankovic, M. Hanson, A. Barth,
J. Lach, and G. Zhou: Accurate, fast fall
detection using gyroscopes and accelerometer-
derived posture information, Sixth Intl.
Workshop on Wearable and Implantable Body
Sensor Networks, pp. 138143 (2009).
[14] Chakeres, I. D., and L. Klein-Berndt: Aodvjr,
aodv simplified, SIGMOBILE Mob. Comput.
Commun. Rev., vol. 6, no. 3, pp. 100101
(2002).
[15] C. E. Perkins, E. M. Belding-Royer, and S. Das:
Ad hoc on-demand distance vector (aodv)
routing, RFC 3561 (Jul. 2003).
[16] S. Sarangi and S. Kar: Performance analysis of
an improved graded precision localization
algorithm for wireless sensor networks, Intl. J.
of Computer Networks & Comm. (IJCNC),
vol. 2, no. 4, pp. 150159 (Jul. 2010).
[17] S. Sarangi and S. Kar: Location estimation with
reactive routing in resource constrained sensor
networks, Intl. Conf. on Sensors and Related
Networks (SENNET09), vol. II, Vellore, India,
pp. 563567 (Dec. 2009).