Abstract— Many mobile ad hoc network protocols use simple flooding, in order to adapt to changes in time varying network topology. Most of the times, a network-wide flood results in redundant packets and increases network congestion, probability of packet collision, low utilization of available bandwidth, and most important, higher power consumption. In this paper, we propose a new cross-layer broadcast scheme to minimize broadcast traffic in mobile ad hoc networks. Our scheme is based on use of received signal strength indicator, RSSI, value to reduce the number of broadcast packets. The effectiveness of the proposed technique is verified using simulations.

Index Terms—Ad hoc networks, AODV, Broadcast, Cross-layer, DSR, IEEE 802.11, RREQ, RSSI

I. INTRODUCTION roadcasting is most widely used technique within mobile ad hoc networks for network-wide information

dissemination. Many protocols for mobile ad hoc networks rely on broadcasting for their successful operation, e.g., on-demand routing protocols flood entire network for route discovery and maintenance [1].

The efficiency of broadcast implementation determines routing protocols scalability, which in turn determines capacity of mobile ad hoc networks. In a simplest broadcast implementation, the entire mobile ad hoc network search space is explored by network-wide flooding of broadcast packets. In general, routing protocols for mobile ad hoc networks, e.g., Ad Hoc On-demand Distance Vector, AODV, and Dynamic Source Routing, DSR, flood the entire network with broadcast packets during route discovery [2, 3, 12].

However, flooding often results redundant packets and increases network congestion, probability of packet collisions and power consumption [6]. Further, in mobile ad hoc networks utilizing IEEE 802.11 MAC, broadcast packets may stymied unicast packets transmissions due to lack of exponential backoff policy for broadcast packets [5]. Therefore, broadcast overhead limits capacity of mobile ad hoc networks, operating in resource constrained environment.

A number of techniques have been proposed to reduce broadcast overhead [7, 8]. Recently cross-layer protocol

designs for mobile ad hoc networks have drawn significant interest among researchers because of their improved performance. The cross-layer design exploits information available on different layers of network stack [7].

In this paper, we propose a new cross-layer approach for the reduction of broadcast traffic in mobile ad hoc networks. In particular, we focus on ad hoc networks where nodes exhibit high degree of mobility, all the times. As a proof of concept, we have implemented and studied our proposed theory on on-demand routing protocol AODV [3, 12]. This allows us to measure effectiveness of the proposed method, when embedded into a particular ad hoc network protocol.

The rest of the paper is organized as follows. Section 2 gives an overview of route discovery process in on-demand routing protocols, followed by proposed technique in detail. Section 3, presents performance evaluation of proposed technique and finally, Section 4 concludes the paper.

II. BROADCAST TRAFFIC REDUCTION This section gives a concise introduction to route discovery

process used in on-demand routing protocols on which we base and study our proposed broadcast reduction framework. After presenting summary of route discovery, we present motivation behind this work and proposed theory in details.

A. Application of Broadcast in on-demand routing protocols In on-demand routing protocols, AODV and DSR, a route

is discovered by flooding route request, RREQ, packets throughout the network [2, 3, 12]. A node in such a protocol after receiving RREQ, replies to originator of RREQ, if it has route in its route cache, otherwise rebroadcasts RREQ further in the network, if it does not have a cached route. This simple flooding often results in redundant RREQ messages and therefore, increases network congestion. In these protocols RREQ broadcast overhead is reduced by utilizing expanding ring search, however, possible only when some node in route request propagation path has the desired route in its route cache [12].

In [9], authors have shown that RREQs dominate AODV’s routing load and routing overhead is substantially higher (about 75 percent) in AODV than in DSR. Further, in the presence of high node mobility, the frequency of route discoveries in AODV is directly proportional to the number of route breaks [9]. Taking these results into account, we have

On the Reduction of Broadcast Traffic in Mobile Ad Hoc Networks

Ashish Shukla Marvell India Pvt. Ltd., Pune 411006

India shukla.a@gmail.com

B

1-4244-1312-5/07/$25.00 © 2007 IEEE 1581

implemented our proposed idea on AODV.

B. Broadcast traffic reduction In mobile ad hoc networks, node mobility results in

frequent communication link failures. In a mobile ad hoc network, where nodes keep moving all the times, routing protocol most of the times performs route discovery. As nodes are moving, a node receiving a packet on the boundary of communication range of the transmitter node may be allowed to drop the packet, as soon receiver may move out of communication range of the transmitter. To approximate the distance between receiver and transmitter, we have used received signal strength information (RSSI) values, received from wireless network interface card.

The IEEE 802.11 standard specifies implementation of RSSI as an optional entity [5]. However, all the wireless network interface cards, to the best of our knowledge, implement this and provide access to observed RSSI values of the received packets, through device driver API calls.

Assuming that all nodes use same transmit power for packet transmission, we have observed that the received signal strength information, RSSI, values of the received packets vary as a function of distance, d, between transmitter and the receiver. The RSSI value, Pr, associated with received packet, follows the following propagation model:

ndtPdrP −= α)( (1)

where Pt represents transmit power used by nodes which is constant and same for all the nodes, d represents distance between receiving node and the node that has transmitted the packet, α is a constant (depends on antenna characteristics, system loss factor and carrier wavelength), and path loss exponent n satisfies 1.6 ≤ n ≤ 6 [11]. Equation (1) does not take into account imperfect channel conditions due to multipath fading and other transient interferences.

As nodes move, the distance between them varies and therefore RSSI, Pr, values. We define a threshold value for Pr, hereinafter called Pthr. When a node receives a broadcast RREQ, it checks received Pr against Pthr. If a RREQ packet is received with Pr less than Pthr , the packet is dropped silently. However, when a RREQ packet is received with Pr more than or equal to Pthr, the packet is processed according to underlying routing protocol.

The value of RREQ drop threshold Pthr shall be decided based on given ad hoc network configuration. In general, the distribution of Pr is of logarithmic nature, and majority of broadcast packets are received with Pr close to minimum value of observed Pr values. Therefore, in many situations, the value of Pthr shall be close to minimum Pr. This value represents the case when receiving node is on the boundary of the communication range of transmitter node.

Dropping broadcast RREQ packets based on Pthr will certainly reduce network traffic but at the same time may degrade underlying protocol performance. Further, in certain

cases, it may even increase network load and congestion due to reactive nature of on-demand routing protocols. We will prove here and also with simulation that the above does not hold true in general, and proposed method decreases broadcast traffic and improves underlying protocol performance.

Assuming at any particular instant of time, there are two moving nodes within communication range of each other with distance between them dt, such that first node has transmitted a RREQ, and second node has received it. This point of time receiving node compares received Pr against Pthr and suppose found it less than Pthr and drops the packet. In the next time instant one of the following three cases will occur assuming the distance between these nodes is now dt+1. We define ∆ = dt+1 - dt. 1) ∆ > 0

In this case nodes are moving away of each other and therefore, probability is high that they will soon move out of communication range of each other, if not already, and dropping RREQ, therefore, will improve underlying protocol. 2) ∆ < 0

This is the case when two nodes are moving close to each other and proposed protocol has been already dropped the received RREQ packet. In this case, the decision of dropping packet was wrong. However, this will not degrade underlying protocol performance as long as node, which has originated RREQ, gets a reply of RREQ from some other node. If the source of RREQ, however, does not get any response of RREQ previously sent, it will rebroadcast RREQ and as nodes are now close to each other, received RREQ will be processed and not dropped. 3) ∆ = 0

To handle such a case, the proposed framework maintains a list of RREQ packets that were dropped by the framework in past. This list is searched and updated with the address of the RREQ sender node and broadcast ID of RREQ. When a packet is dropped it is entered into this list and in future when a node receives same RREQ that is present in this list, it forwards the received RREQ to underlying protocol and delete the RREQ entry from dropped packet list. Dropped packet list entries should be kept for at least the time, it takes a message to traverse entire mobile ad hoc network. This mechanism ensures that in case a node accidentally drops the RREQ packet, it should not repeat the same mistake in future.

III. PERFORMANCE EVALUATION We have used simulations, using ns-2.29 with CMU

Monarch wireless extensions, to study effectiveness of our proposed framework [4]. All simulation results were obtained without dropped packet list implementation. In the following, we describe the simulation environment and the performance metrics used, followed by simulation results.

A. Simulation environment We executed the experiments with ns-2.29 network

simulator with CMU Monarch wireless extensions [4]. All simulations carried out with 50 nodes and simulated 1000

1-4244-1312-5/07/$25.00 © 2007 IEEE 1582

seconds of real time. The radio model for each node was similar to Lucent Technologies WaveLAN product, which is a shared-media radio with a raw capacity of 2 Mbit/s and a 250m nominal communication range. The distributed coordination function (DCF) protocol of the IEEE 802.11 for wireless LANs was used as the MAC layer [5]. Link or node failure was detected by the feedback from the MAC layer. In case of failing to deliver a packet to the next hop, the MAC layer informs the routing layer of the failure. The nodes communications were modeled using constant bit rate (CBR) traffic sources sending data in 512 byte packets at a rate of four packets/s. A total of 20 CBR connections were used, with each node being the source of zero, one, or two connections, and a total of 14 different originating nodes.

Random Waypoint mobility model was used and nodes were spread in a square region, having dimension of 670m x 670m. In this model each node picks a random location in the simulated area and proceeds there at a speed chosen uniformly from [1, 20]m/s [10]. When the node reaches this point, it remains stationary for a fixed time --pause time--, after that picks another location to move to and repeats the cycle. Pause time controls the degree of mobility by making each node remain stationary for a definite period of time, before it moves to the next position. The pause time was kept low in our simulation, therefore, allows us to study the behavior of the proposed mechanism under high node mobility.

We have performed all the studies with two thresholds Pthr equal to 6.0e-10 and Pthr equal to 1.0e-09. We have implemented our scheme as a MAC layer filter driver, which dropped all the RREQ packets received with Pr less than Pthr. All packets that were received with Pr more than or equal to Pthr, were passed to underlying routing protocol for further processing. The selection of Pthr was done by studying the distribution of received Pr values of RREQ packets.

B. Performance metrics We have used standard metrics to evaluate performance of

routing protocol and broadcast traffic reduction. The performance of proposed framework is evaluated on the following metrics: 1. Average end-to-end Delay: total delay experienced by

data packets including interface queuing delays, route discovery delays, retransmission delays at the MAC, propagation delays and transfer delay.

2. Packet delivery ratio: ratio of the packets delivered to the final destination to those generated by the CBR sources.

The above two metrics capture overall routing protocol performance including end-to-end delay and loss rate seen by higher layer protocols, in the presence of proposed framework. However, to measure effectiveness of proposed framework on broadcast traffic reduction, the following metrics are used. 3. Total RREQ transmissions: total number of RREQ

transmissions for a given simulation run. This metric measures network congestion. A low value of this metric is desired.

4. Total MAC collisions: defined as total number of packet collisions at MAC layer. This is an ideal metric to capture MAC overhead including power consumption due to MAC retries. Smaller value of this metric is desired.

C. Simulation results We have performed simulations under high mobility

scenarios. First, we present results of routing protocol performance, followed by broadcast traffic metrics. In all of the following figures, 'AODV-R1' refers to all-optimized standard AODV protocol incorporating our broadcast reduction mechanism with Pthr set to 6.0e-10. Similarly, 'AODV-R2' refers to AODV protocol with Pthr set to 1.0e-09. The results obtained with modified AODVs are compared with original AODV protocol. Each data point in simulation results represents an average of ten runs.

Fig. 1, presents average end-to-end delay. The proposed framework significantly improves average end-to-end delays when node mobility is high. This is because of frequent link breaks due to node movements. In general, AODV-R1 performs on average more than 16 percent better than the original AODV. However, AODV-R2 does not give better results over AODV-R1. As Pthr increases, in AODV-R2, more RREQ packets are dropped. This may result in longer path and therefore, may increase average end-to-end delay. Fig. 2, presents packet delivery ratio of AODV-R1, AODV-R2 and original AODV. The proposed broadcast reduction mechanism improves packet delivery ratio of AODV-R1 and AODV-R2. The improved average end-to-end delay and packet delivery ratio are due to reduced multiple access interferences and network congestion.

Fig. 3, gives total RREQ transmissions per simulation run. The modified AODV-R1 and AODV-R2 generate lesser number of RREQ requests than the original AODV. In general, AODV-R1 reduces broadcast traffic on average by more than 18 percent, whereas AODV-R2 reduces broadcast traffic by more than 22 percent. Fig. 4, gives number of collisions experienced by MAC layer. In general, AODV-R1 reduces MAC collisions by more than 20 percent, whereas AODV-R2 reduces MAC collisions by more than 30 percent. In summary, proposed broadcast reduction mechanism significantly reduces network congestion and reduces power consumption.

We have observed that when Pthr is close to minimum value of received Pr values, routing protocol performance increases while network traffic decreases. This is the ideal case when nodes are on the boundary of the communication range of each other. However, as Pthr increases, due to reactive nature of the underlying routing protocol, the network traffic may even increase. It is observed that increasing Pthr value does not improve protocol performance. The idea of maintaining dropped packet list per node can be used in such cases.

IV. CONCLUSION In this paper we have presented a very simple yet powerful

technique for reducing broadcast traffic and therefore power

1-4244-1312-5/07/$25.00 © 2007 IEEE 1583

consumption in mobile ad hoc networks. The proposed technique can be implemented as a simple MAC layer filter driver, and has zero overhead. The simulation results show significant performance improvement for the broadcast traffic reduction and underlying protocol performance. The proposed technique can be easily generalized to other ad hoc network protocols. Deriving optimal value of Pthr and performance study with more number of nodes and network traffic are part of future work. We would also like to do experiments under noisy, fading and multipath effects, channel conditions.

REFERENCES [1] Internet Engineering Task Force MANET Working Group, Mobile Ad

Hoc Networks (MANET) Charter. Available at http://www.ietf.org/html.charters/manet-charter.html.

[2] David B. Johnson, David A. Maltz, and Josh Broch, “ The Dynamic Source Routing Protocol for Multihop Wireless Ad Hoc Networks,” In Ad Hoc Networking, edited by Charles E. Perkins, chapter 5, pages 139–172, Addison-Wesley, 2001.

[3] Charles E. Perkins and Elizabeth M. Royer, “Ad-Hoc On-Demand Distance Vector Routing,” In Second 1EEE Workshop on Mobile Computing Systems and Applications, pages 90-100, February 1999.

[4] The Network Simulator –ns2, online

http://www.isi.edu/nsnam/ns/ [5] ANSI/IEEE Std 802.11, 1999 Edition. IEEE Standards for Information

Technology - Telecommunications and Information Exchange between systems – Local and Metropolitan Area Network – Specific Requirements. Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications.

[6] S. Ni, Y. Tseng, Y. Chen, and J. Sheu, “The broadcast storm problem in a mobile ad hoc network,” Proceedings of ACM/IEEE Int. Conf. on Mobile Computing and Networking (MOBICOM), 1999, pp. 151–162.

[7] Kang I. and Poovendran R., "Scalable Power-Efficient Broadcast Over Densely Deployed Wireless Ad Hoc Networks," in Proceedings of IEEE Wireless Communication and Networking Conference (WCNC), 2005, pp. 2063-2068.

[8] Dai F. and Wu J., "Performance Analysis of Broadcast Protocols in Ad Hoc Networks Based on Self-Pruning," IEEE Transactions on Parallel and Distributed Systems, VOL. 15, NO. 11, NOVEMBER 2004, pp. 1027-1040.

[9] C. E. Perkins, E. M. Royer, S. R. Das, and M. K. Marina, "Performance Comparison of Two On-Demand Routing Protocols for Ad Hoc Networks," IEEE Personal Communications, February 2001, pp. 16-28.

[10] J. Yoon, M. Liu and B. Noble, "Random waypoint considered harmful," in Proceedings of IEEE Infocom 2003, San Francisco (CA), April 2003.

[11] Rappaport T.S., Wireless Communications Principles and Practice. Upper Saddle River, NJ, USA: Prentice Hall PTR, 2nd ed., 2002, ISBN 0-13-042232-0.

[12] C. Perkins, E. Belding-Royer, S. Das, Ad hoc On-Demand Distance Vector (AODV) Routing, Internet experimental RFC 3561, July 2003.

0

20

40

60

80

100

0 50 100 150 200 250

Ave

rage

del

ay (m

s)

pause time (secs)

AODVAODV-R1AODV-R2

Figure 1. Average end-to-end delay

0.8 0.82 0.84 0.86 0.88

0.9 0.92 0.94 0.96 0.98

1

0 50 100 150 200 250

Pack

et d

eliv

ery

ratio

pause time (secs)

AODVAODV-R1AODV-R2

Figure 2. Packet delivery ratio

x 105

1

1.5

2

2.5

3

3.5

4

4.5

0 50 100 150 200 250

Tota

l RR

EQ tr

ansm

issi

ons

pause time (secs)

AODVAODV-R1AODV-R2

Figure 3. Total RREQ transmissions

x 104

0

5

10

15

20

0 50 100 150 200 250

Tota

l MA

C c

ollis

ions

pause time (secs)

AODVAODV-R1AODV-R2

Figure 4. Total MAC collisions

1-4244-1312-5/07/$25.00 © 2007 IEEE 1584