Fuzzy-controlled Scheduling of Real Time Data Packets (FSRP) in Mobile Ad Hoc Networks

Dr. Anuradha Banerjee et al, International Journal of Computer Science and Mobile Computing, Vol.5 Issue.5, May- 2016, pg. 507-513
© 2016, IJCSMC All Rights Reserved 507
Available Online at www.ijcsmc.com
International Journal of Computer Science and Mobile Computing
A Monthly Journal of Computer Science and Information Technology
ISSN 2320–088X
IMPACT FACTOR: 5.258
IJCSMC, Vol. 5, Issue. 5, May 2016, pg.507 – 513
Fuzzy-controlled Scheduling of
Real Time Data Packets (FSRP)
in Mobile Ad Hoc Networks
Abstract—In ad hoc networks, the data packets can be mainly classified into two categories – real time and
non-real time. Definitely the scheduling of real time data packets is going to be different from non-real time
ones. The deadline of delivery is a very important criterion to be met in case of real time packets. The other
requirements like velocity and energy concerns are enforced by the inherent dynamic nature of ad hoc
networks. Sometimes it becomes completely unnecessary to forward a real time data packet when the deadline
is impossible to meet or when the destination is on the verge of death. Also if the source nodes start sending
real time packets at a very high rate, then it will not only increase congestion in the network but also hamper
the forwarding of non real time packets since non real times have priority lesser than the real time packets.
The present algorithm FSRP is designed considering all these factors and the performance improvement it
produces is very significant compared to other state of the art scheduling protocols.
Keywords—Ad hoc networks; delay; real time, route-request; scheduling.
Introduction
An ad hoc network is a collection of wireless nodes which form a temporary network without relying on an
existing infrastructure or centralized administration. These networks are deployed mainly in emergency situations
like battlefield, natural disasters like earthquake, floods etc. [1-5]. Many routing protocols have been proposed in
ad hoc networks so far. In all of them, when the destination node is out of the radio-range of the source node the
communication has to be multi-hop where some nodes act as router to bridge the gap between the source and
destination nodes [6- 10]. If a node receives multiple message forwarding requests, it serves one of them and
stores the others in its queue. The order in which these requests will be served is termed as a schedule. The job of
a scheduler is to pick up that particular schedule that is expected to produce the best performance.
Dr. Anuradha Banerjee Dr. Paramartha Dutta Abu Sufian
Dept. of Computer Applications Dept. of Computer and Systems Science Dept. of Computer Science
Kalyani Govt. Engg. College Visva-Bharati University, Santiniketan University of Gour-Banga
West Bengal, India West Bengal, India West Bengal, India
anuradha79bn@gmail.com paramartha.dutta@gmail.com sufian.asu@gmail.com

The choice of scheduling algorithm has a significant effect on the overall performance of the route, especially
when the traffic load is high [10-14]. A packet scheduling scheme and algorithm called RACE [16] for real-time
large scale networks was proposed which uses Bellman-Ford algorithm in order to find out ways with less traffic
and delay. Earliest Deadline First (EDF) scheduling algorithm was used in RACE to transmit packets with
shortest deadline. [16] presents the mostly used operating system of Wireless Network and differentiate them as
Cooperative and Preemptive. Co-operative scheduling algorithms are based on Adaptive Double Ring Scheduling
(ARDS) and EDF [10, 16], that has two queues with various priorities. Based upon the deadline of the arriving
packets, the scheduler switches between the two queues. Cooperative schedulers are used in applications with
limited resources. Preemptive Scheduling is based on EFRM scheme which is an extension of Rate Monotonic
(RM) scheme. Dynamic multilevel priority (DMP) [16] is a scheduling algorithm that uses three different priority
queues – a) real-time (Priority-1) b) non-real time data packets from lower level nodes (Priority-2) c) non-real
time data packets present at the node itself (Priority-3). Shortest Job First (SJF) scheduler is used to process the
non-real time data packets that are present at the same level priority. Time-Division Multiplexing Access is
applied for the processing of data packets at different levels. For example, nodes that are situated at the lowest
level and the immediate next lowest level can be allotted timeslots 1 and 2 respectively.
To the best of the author’s knowledge, as far as the scheduling of real time data packets is concerned, there
does not exist a scheme in the literature for ad hoc networks which consider the stability of routes and energy of
the routers along with source and destination nodes. The present article focuses on proposing one scheduling
scheme that will prioritize the real time message packets based on deadline, relative velocity between consecutive
routers in the path along with their residual energies. The reason is that if a path is not stable enough in terms of
energy and velocity, then real time packets should be urgently forwarded through those paths. Reason is that if the
path is broken then a huge amount of time and energy will be wasted in the route discovery process to build a new
path from that router to the destination. As a result, the packet may miss its deadline which is not at all affordable.
FSRP in Detail
As soon as a real time message packet generated by source ns for destination nd arrives at router ni at time t, ni
first checks whether it is really possible to forward the data packet to the destination within the deadline. If it is
impossible then the packet is dropped by ni.
Let D(Ps,d, i) be the time span required for a data packet generated at source ns to travel to destination nd from
router ni along the path P. P is the current path being followed by ns to transmit packets to nd in real time. Also
assume that t is the current time and the real time packet which arrived at ni, needs to be delivered to nd within
time td. If ( t+ D(Ps,d, i)) > td then the packet cannot be delivered to the destination within its stipulated time. As a
result, the packet is dropped by ni. According to the discharge curve of batteries heavily used in ad hoc networks,
at least 40% of total charge is required to remain in operable condition. If at time t, the destination nd is equipped
with less than 40% of its total charge, then the current real time packet for nd is dropped by ni. Otherwise ni
forwards it. Decision to pick up one particular packet is taken by the real time fuzzy controller (RTF) which is
embedded in every node. Like DMP, the scheduling process in FSRP is also non-preemptive.
Design of Real-time Fuzzy Controller (RTF)
The input parameters are deadline impact (LI), stability impact (SI) and source priority (SP). These are
mathematically formulated in subsection A and combined in fuzzy rule bases in subsection B in this section.
Input Parameters of Real Time Fuzzy Controller (RTF)
The input parameters are described below:
i) LI(Ps,d, i) = (D(Ps,d, i)+t) / td (1)
LI(Ps,d, i) specifies the ratio of the estimated time at which the real time packet at router ni is supposed to
reach the destination nd and the specified deadline td within which the packet must reach the destination.
Definitely LI(Ps,d, i) is less than or equal to one otherwise the packet would have been dropped. So the value
of LI lies between 0 and 1. Higher the value of LI, higher is the urgency of forwarding the packet since the
estimated time of arrival of the real time packet to nd is very close to its deadline.
ii) SI(Ps,d) = 1 – [ ∑ relvj, j+1(t) engj, j+1(t) ]0.5
/ h(Ps,d) (2)
nj, nj+1 Ps,d
Formulation of relvj, j+1(t) is illustrated based on figures 1a, 1b and 1c. nj and nj+1 are two consecutive routers in
Ps,d.

Vj+1(t)
Vj(t)
Vj+1(t)
Vj(t)Vj(t)
Vj+1(t)
Vj(t)
 > 90  = 90
Fig 1a: The velocities make an obtuse angle Fig 1b: The velocities make right angle
 < 90
Fig 1c: The velocities make an acute angle
For the situation in figure 1a
Horizontal component of relative velocity relHj,j+1(t) = vj+1(t) cos(180 - ) + vj(t) = vj+1(t) cos() + vj(t)
Vertical component of relative velocity relVj,j+1(t) = vj+1(t) sin(180 - ) = vj+1(t) sin()
For the situation in figure 1b
Horizontal component of relative velocity relHj,j+1(t) = vj(t)
Vertical component of relative velocity relVj,j+1(t) = vj+1(t)
For the situation in figure 1c
Horizontal component of relative velocity relHj,j+1(t) = |vj+1(t) cos() - vj(t)|
Vertical component of relative velocity relVj,j+1(t) = vj+1(t) sin()
relvj, j+1(t) = {relHj,j+1(t) (1+ relVj,j+1(t))}0.5
/ (1+vj(t) + vj+1(t)) (3)
relvj, j+1(t) specifies the relative velocity component of the link from nj to nj+1 at time t. Similarly engj, j+1(t)
indicates the energy component of the same link at the same time. Assuming that ej(t) is the consumed battery
power of nj till time t and Ej is its total battery power, engj, j+1(t) is formulated in (4).
engj, j+1(t) = {(ej(t) /Ej) (ej+1(t) /Ej+1)}0.5
(4)
Lesser the value of relvj, j+1(t) lesser is the chance that the link between nj and nj+1 will break soon in near future as
a consequence of their movements. Similarly, lesser the value of engj, j+1(t) less will be charge consumed by the
involved nodes and high will be their remaining energies. As a result, less chance is there that the nodes will die
soon or the link will break soon as a result of battery exhaustion of the two involved nodes nj and nj+1. h(Ps,d) is
the number of hops in the path from ns to nd along the path P. Energy and relative velocity related components
(eng and relv) have been averaged based on per hop basis.
From the formulation (2) it is evident that if the value of SI is small, there is high chance of link breakage in
Ps,d due to high relative velocity and/or scarcity of energy in routers and so, the real time packet should be sent
urgently through this path. Please note that SI lies between 0 and 1.

iii) RYs,i(t) = pks,i(t) /(∑ pkm,i(t) ) (5)
nmFi(t)
RYs,i(t) specifies the priority of the source ns to router ni at time t. The priority decreases if ns has already
transmitted a huge number of real time packets to ni for forwarding. The reason is that real time packets impose
more thrust on the router to do its processing before the others. Therefore it also hampers forwarding of other
packets. So, real time packets should not be transmitted in huge numbers unless it is utmost necessary. Fi(t) is the
set of all sources that transmitted real time packets to ni for forwarding, till time t. pkm,i(t) is the number of real
time packets transmitted by node nm to ni for forwarding till time t. From the formulation in (5) it is evident that
RYs,i(t) lies between 0 and 1. Values close to 0 increase the priority of ns to ni.
Rule Bases of RTF
All three of LI, SI and RY range between 0 and 1. They are uniformly divided into crisp ranges. 0-0.25 is
indicated as fuzzy premise variable a, 0.25-0.50 as b, 0.50-0.75 as c and 0.75-1.00 as d. Table 1 presents the fuzzy
combination of LI and SI producing temporary output t1.
Table 1: Fuzzy Combination of LI and SI producing t1
LI 
SI
a b c d
a b c c d
b b b c d
c a a b c
d a a a a
High values of t1 correspond to the situation where LI is very high (d) and SI is very low (a). On the other
hand low values of t1indicate the situation when LI is very low (a) and SI is very high (d). None of these
parameters completely dominate one another. Table 2 shows the fuzzy combination of t1 and RY. It is quite
practical that t1 dominates RY since t1 is the combination of two parameters (LI is concerned with the
urgency with respect to the pre-specified deadline whereas SI is concerned with the fragility of the path i.e.
how easily it can break; its good if the real time packets can be delivered through the path before it breaks
otherwise a new route will have to be discovered that will greatly increase the cost of control messages
injected into the network) both of which strongly contribute to determine the urgency of forwarding the
message. High values of t1 contribute to increase urgency of forwarding whereas low values of RY encourage
the router to urgently forward the message. The output produced by table 2 is UR. It is quite understandable
that its value also lies between 0 and 1. It follows the same range distribution as the input parameters of RTF.
Table 2: Fuzzy Combination of t1 and RY producing t1
t1 
RY
a b c d
a b c d d
b b c d d
c b b c d
d a b b c
Simulation Results
Simulation of the mobile network has been carried out using ns-2 [15] simulator on 800 MHz Pentium IV
processor, 40 GB hard disk capacity and Red Hat Linux version 6.2 operating system. Graphs appear in figures
2, 3 and 4 showing emphatic improvements in favor of FSRP. Number of nodes has been taken as 20, 50, 100,
150 and 200 in different independent simulation studies. Speed of a node is chosen as 5m/s, 10 m/s, 25 m/s, 35
m/s and 50 m/s in different simulation runs. Transmission range varied between 10m and 50m. Used network
area is 500m 500m. Used traffic type is constant bit rate. Mobility models used in various runs are random
waypoint, random walk and Gaussian. Performance of the protocols DMP-AODV, DMP-ABR and DMP-FAIR
(DMP embedded versions of the protocols AODV, ABR and FAIR) are compared with their FSRP embedded
versions FSRP-AODV, FSRP-ABR and FSRP-FAIR respectively. In order to maintain uniformity of the
implementation platform, we have used ns-2 simulator for all the above-mentioned communication protocols.

The relevant simulation matrices are per packet (real time) waiting time per router in seconds vs number of
nodes, percentage of real time packets that could be successfully delivered to their own destinations vs number
of nodes and per node per session message cost vs number of nodes. Simulation time was 1000 sec. for each
run.
Figure 2: Graphical illustration of per packet waiting time per router vs number of nodes
Figure 3: Graphical illustration of percentage of packets that could be successfully delivered to the destination vs number of nodes
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
PERPACKETWAITINGTIMEPER
ROUTER
20 50 100 150 200
NUMBER OF NODES
DMP-AODV
FSRP-AODV
DMP-ABR
FSRP-ABR
DMP-FAIR
FSRP-FAIR
0
20
40
60
80
100
120
PERCENTAGEOFPACKETDELIVERY
20 50 100 150 200
NUMBER OF NODES
DMP-AODV
FSRP-AODV
DMP-ABR
FSRP-ABR
DMP-FAIR
FSRP-FAIR

Figure 4: Graphical illustration of per node per session message cost vs number of nodes
Experimental Results
As the number of nodes continues to increase, more real time packets arrive at routers. If the scheduling
method is not efficient, a lot of delay will be unnecessarily produced by the system where the average waiting
time will increase. Although DMP [16] performs better than EDF , still it does not consider the factors like energy
and relative velocity. The real time packets with short deadlines are prioritized and this is done in RTF through
the input parameter SI. Also the packets for which deadline cannot be maintained, are unconditionally dropped by
the router who discovers it first. This greatly saves the cost of messages in the network. FSRP estimates urgency
to forward a packet based on its deadline, relative velocity between routers in the path being followed by the
packet and the residual energies of those routers. If the residual energy of any one router in the path is poor, then
the path may break up soon and for that a new route discovery session needs to be initiated by the current router.
This results in the injection of a huge number of messages in the system increasing message cost. As a result,
packet loss will increase due to severe congestion and packet collision in the network. FSRP considers all these
aspects for which it produces much lesser message cost and high packet delivery ratio compared to DMP
embedded versions of the state-of-the-art routing protocols in ad hoc networks like AODV, ABR and FAIR. This
is illustrated in figures 3 and 4.
Due to the significant improvements in terms of message cost, delay in each router also decreases. The reason
is that, if a route-request packet is being processed by a router and at that time a real time packet comes, then the
packet will have to wait even if it has a very short deadline or travels through an unstable path in both DMP and
FSRP due to the non-preemptive nature of both the protocols. But since the number of route-requests is much less
in FSRP compared to DMP, the waiting time per router decreases for each packet. The non-preemptive nature of
DMP is preserved in FSRP in order to minimize the number of context switches as much as possible because
context switching has got its own overhead.
Although FSRP embedded versions of the protocols AODV, ABR and FAIR perform much better than their
DMP embedded versions still the improvement is the maximum for AODV and minimum for FAIR. The reason
is that FAIR as a protocol already consider the energy and relative velocity of routers to select the best path,
although the formulations are much different compared to FSRP. So, FAIR, in rear cases, communicates through
unstable paths.
CONCLUSION
Scheduling of real time data packets is a very important part in communication in ad hoc networks. If the
average waiting time of these packets is decreased then automatically per node waiting time will decrease with
great increase in the network throughput or packet delivery ratio. Unlike the previous scheduling algorithms,
FSRP rightly considers the stability of the path in terms of relative velocity of the routers and their residual
energies. Moreover, real time packets that cannot be delivered in time to the respective destinations are
unconditionally dropped by the first router that discovered this. All these tremendously contribute to improve
the performance of the underlying routing protocols.
0
100
200
300
400
500
600
700
PERNODEPERSESSIONMESSAGECOST
20 50 100 150 200
NUMBER OF NODES
DMP-AODV
FSRP-AODV
DMP-ABR
FSRP-ABR
DMP-FAIR
FSRP-FAIR

REFERENCES
[1] C.E. Perkind & P. Bhagat,”Highly Dynamic Destination Sequenced Distance Vector Routing (DSDV) For Mobile Computers”
Computer Communications Review, vol. 24 no. 4 1994, pp. 234–244.
[2] S. Murphy, J.J. Garcia-Luna-Aceves, “An Efficient Routing Protocol for Wireless Networks”, ACM Mobile Networks and
Applications Journal, pages 183-197, Nov.1996. http://citeseer.nj.nec.com/10238.html
[3] T.W. Chen, M. Gerla, “Global State Routing: A new routing scheme for wireless ad hoc networks” , IEEE Conference On
Information, Communication and Control, 1998
[4] C.C. Chiang et. Al.,”Routing in clustered multi-hop wireless networks with fading channel”, IEEE Conference on Innovative Systems.
1997, Singapore
[5] Broch, J., Maltz, D., Johnson, D., Hu, Y., , And Jetcheva, J. A performance comparison of multi-hop wireless ad hoc network routing
protocols. In Proceedings of the Fourth Annual ACM/IEEE International Conference on Mobile Computing and Networking
(MobiCom ’98) (Dallas, Texas, USA, Aug. 1998).
[6] A. Kodole, P.M. Agarkar, “A survey of routing protocols in mobile ad hoc networks”, Multi-disciplinary Journal of Research in Engg.
& Tech, vol.2 no. 1, pg – 336-341, 2015
[7] C.E. Perkins, E.M. Royer, “Ad Hoc On-demand Distance Vector Routing” IEEE Wokshop on Mobile Computing Systems and
Applications., 1999
[8] T.X. Brown, H.N. Babow & Q. Zhang, “Maximum Flow Life Curve for Wireless Ad Hoc networks” ACM Symposium on Mobile Ad
Hoc Networking and Computing, USA, 2001
[9] C.K. Toh & P. Bhagwat, “Associativity based routing for mobile ad hoc networks”, Wireless Personal Communications, 1997, vol. 4
no. 2., pp. 1-36
[10] A. Banerjee, P. Dutta, “Fuzzy-controlled adaptive and intelligent route selection (FAIR) in ad hoc networks”, European Journal of
Scientific Research, vol. 45 no. 3, 2010, pp. 367-382
[11] D.R.S. Victoria, S.S. Kumar,”Efficient bandwidth allocation for packet scheduling”, International Journal of Future Computing and
Communications, vol. 1 no. 4, Dec 2012
[12] C. Annadurai, “Review of Packet Scheduling Algorithms in Mobile Ad Hoc Networks”, International Journal of Computer
Applications, vol. 15 no. 1, Feb 2011
[13] A. Banerjee, P. Dutta, “Delay-efficient Energy and Velocity Conscious Non-Preemptive Scheduler (DEV-NS) For Mobile Ad Hoc
Networks”, International Journal of Advanced Networking and Applications, vol. 5 issue 4, 2014
[14] R. Parasher, Y. Rathi, “A-AODV: A modern routing algorithm for mobile ad hoc networks”, International Research Journal of Engg.
And Technology, vol. 2 issue 1, 2015
[15] http://www.isi.edu/nsnam/ns/tutorial/
[16] R.A. Kumar, K.A. Varshini, Multilevel Priority Packet Scheduling Scheme for Wireless Networks, International Journal of Distributed
and Parallel Systems (IJDPS) Vol.5, No.1, May 2014

Fuzzy-controlled Scheduling of Real Time Data Packets (FSRP) in Mobile Ad Hoc Networks

More Related Content

What's hot (17)

Similar to Fuzzy-controlled Scheduling of Real Time Data Packets (FSRP) in Mobile Ad Hoc Networks (20)

Recently uploaded (20)

Fuzzy-controlled Scheduling of Real Time Data Packets (FSRP) in Mobile Ad Hoc Networks