AODV-Based Backup Routing Scheme in Mobile Ad Hoc Networks

Tsung-Chuan Huang, Sheng-Yu Huang and Lung Tang Department of Electrical Engineering National Sun Yat-Sen University

Kaohsiung, Taiwan tch@mail.nsysu.edu.tw, {m963010059, d953010007}@student.nsysu.edu.tw

Abstract—As effective routing is critical in mobile ad hoc networks (MANETs), the Ad hoc On-demand Distance Vector (AODV) has been extensively studied in recent years. Given that AODV requires a new route discovery procedure whenever a link breaks, such frequent route discoveries incur a high routing overhead and increase end-to-end delay. Therefore, by modifying the AODV protocol, this work presents a novel backup routing scheme capable of repairing disrupted links locally without activating a route re-discovery procedure. Additionally, backup paths are established based on 2-hop neighbor knowledge. These backup paths are geographically close to the primary path in order to provide efficient recovery from route failure and maintain an adequate routing length. Simulation results indicate that the proposed backup routing scheme obtains a lower average end-to-end delay and less routing overhead than those of the Ad hoc On-demand Multipath Distance Vector (AOMDV) and the conventional AODV.

Keywords-MANET; AODV; multipath routing; backup route

I. INTRODUCTION A mobile ad hoc network (MANET) is a flexible, self-

organizing wireless network consisting of a collection of wireless mobile nodes without a centralized control or infrastructure. Mobile nodes in such a network are capable of moving independently and communicating with other nodes using direct wireless links or multi-hop wireless links through a series of intermediate nodes. This network architecture is very useful in such environments as highly dynamic topologies without base stations. Some examples of MANET applications include battlefields, emergency services, moving vehicles, and conference rooms.

However, MANET characteristics, such as a dynamically changing topology, absence of a central coordinator and bandwidth-constrained wireless links, make its routing more complex than that of conventional wired networks. Generally, MANET routing protocols can be classified as proactive, reactive, and hybrid, based on the routing information update mechanism. Reactive routing protocols establish routes only when routes are needed by a source node; each node maintains individual routing information to destinations, but does not have a full topological view of the network. The reactive routing protocols typically minimize the number of required broadcasts by only creating routes on demand, as opposed to maintaining up-to-date routing information from each node to all other nodes in the

network as in proactive routing. The Ad hoc On-demand Distance Vector (AODV) [1], one of the most popular reactive routing protocols, offers quick adaptation to dynamic link conditions and low network utilization [2, 3]. However, AODV is a single path routing that requires a new route discovery procedure whenever a link breaks along the route. Such frequent route discoveries result in a high routing overhead and increase end-to-end delay.

Multipath routing increases network resilience to link failure. Multiple paths are either end-to-end disjoint paths or braided paths [4]. For end-to-end disjoint paths, multiple paths are constructed between source and destination nodes where backup paths do not intersect the primary path. Conversely, braided paths typically have no completely disjointed paths from a source to destination, but many partially disjoint backup paths; these backup paths are not independent of the primary path. Recent studies demonstrated that braided paths are more resilient and energy-efficient than end-to-end disjoint paths [5, 6].

This work revises the AODV protocol to provide braided paths structure. The 2-hop neighbor knowledge is utilized to establish backup paths. These backup paths are geographically close to the primary path in order to provide efficient recovery from route failure and reduce the number of route discovery procedure.

The remainder of this paper is structured as follows. Section II introduces related work. Section III details our proposed scheme. Simulation results are shown in Section IV. Finally, conclusions are drawn in Section V.

II. RELATED WORK The AODV is inherently a distance vector routing

protocol that has been optimized for ad hoc wireless networks. When a source node wants to communicate with a destination node whose route is unknown, a path discovery process is initiated to locate the destination node. The source node broadcasts a route request (RREQ) packet to its neighbors, which then forward the RREQ to their neighbors. The forwarding process continues until either the destination or an intermediate node that has a route to the destination node is located. When a RREQ reaches the destination node, the destination node responds by unicasting a route reply (RREP) back along the path in the reverse direction. As the RREP routes back to the source node, the route from the source node to the destination node is established. In the

2010 International Conference on Communications and Mobile Computing

978-0-7695-3989-8/10 $26.00 © 2010 IEEE

DOI 10.1109/CMC.2010.313

254

route maintenance process, when a nodefailure, it generates a route error (RERRERR packet is propagated over routes, wcorresponding routes simultaneously. Whenis sent back to a source node, the source nodroute discovery procedure.

To facilitate multipath support in the Anumber of extensions have been developed.[7] developed an extension for AODV, cademand Multipath Distance Vector (AOMprovides loop-free and disjointed alternatAOMDV protocol, each recipient node reverse routes while processing RREQ packmultiple neighbors. Similarly, one or more Rgenerated via a loop-free path by the destinanode that has a route to the destination nodeach received RREQ packet. These RREPreceived from the source or intermediate creation of multiple forward routes leadidestination node. However, with distaprotocols, the topology information a nodfurther limited. Thus, constructing disjointefrom a source to a destination is difficult.

III. AODV-BASED BACKUP ROUTIN

This work modified the conventional Aenable discovery of backup paths from destination (Fig. 1). The 2-hop neighboutilized to generate backup paths during routdestination selects a route depending on whmaximum number of backup paths. In the pwhen the primary route fails, the conrecovered utilizing these backup paths.

The AODV-based backup routing schemcomponents: local connectivity managemenand path maintenance. The following dethese three components in detail.

A. Local connectivity management In the conventional AODV protocol, a n

local connectivity information by broadmessages periodically. Each node periodiHELLO messages to inform its neighborsmoved away. When a node receives a HELLa neighbor, a route to this neighbor is onrouting table when the neighbor does not the neighbor exists, its lifetime is increnetwork topology changes and HELLO mreceived for a defined period of time, the rou

This work modified the HELLO messaghop neighbor knowledge. Each node perioda HELLO message containing a list of all reach in one hop. When a node receives a it updates its local routing table with the information. Nodes that cannot reach neighdirectly can learn about 2-hop neighbors frthat sent the HELLO message. After a node rmessage from all of its neighbors, it has

e detects a link RR) packet. This while invalidating n a RERR packet de initiates a new

AODV protocol, a . Marina and Das alled Ad hoc On-MDV) routing. It te paths. In the creates multiple ets received from

RREP packets are ation node or any de in response to P packets, when nodes, result in

ing to the same ance-vector-based de can obtain is ed alternate paths

NG SCHEME ODV protocol to

a source to a or knowledge is ute discovery. The hich route has the proposed scheme, nnection can be

me has three main nt, path discovery escribes each of

node may provide dcasting HELLO ically broadcasts s that it has not LO message from nly added to the already exist. If

eased. When the messages are not

ute expires. ge to generate 2-dically broadcasts

neighbors it can HELLO message, HELLO message

hbors in two hops rom the neighbor receives a HELLO 2-hop neighbor

knowledge. In the proposed schemneighbor knowledge to generate badiscovery and maintain up-to-date k

B. Path discovery In this proposed scheme, back

primary path to establish a braidprimary path and backup paths arroute request phase; this work rAODV protocol in this procedure.the primary path, the node next todefined as the border node and the nnode is defined as the reverseadvantage of 2-hop neighbor kbelonging to the primary path canwith its upstream node and revebackup paths are geographically cloestablish a braided paths structure.

To implement the modified AODof the RREQ messages of the origaccordingly (Fig. 2(a)). Two addireverse_border and backup_counoriginal RREQ message. The revethe reverse border node of the primary path. The backup_count fibackup paths along the primary pato the current node. The format ofthe original AODV is also modifiedAn additional field, called the bordthe original RREP message. The border node of the current node Besides, the format of the routing tAODV is modified (Fig. 2(c)). Twbackup_count field of RREQ and thare appended to the original routing

When a source node requires a rsource node broadcasts a RREQ discovery. An intermediate node reRREQ sets up a reverse route to thestep as in the conventional AODV,route using the RREQ that arrives fconventional AODV, the intermed

Figure 1. Structure of AODV-based b

me, each node uses 2-hop ackup paths during route knowledge of such links.

kup paths intersect the ded paths structure. The re established during the revises the conventional When a node exists on

o its downstream node is node next to its upstream

e border node. Taking knowledge, each node n compute backup paths rse border node. These

ose to the primary path to

DV protocol, the format ginal AODV is modified itional fields, called the t, are appended to the

erse_border field records current node along the ield sums the number of ath from the source node f the RREP messages of d accordingly (Fig. 2(b)). der field, is appended to border field records the along the primary path. able entry of the original wo additional fields, the he border field of RREP

g table entry. route to a destination, the

packet to initiate route eceiving the first arriving e source. This is the same , which sets up a reverse first. However, unlike the diate node computes the

backup routing scheme.

255

backup path using the reverse_border fieand the 2-hop neighbor table.

The intermediate node analyzes the neipath that can bypass the upstream node toreverse border node is a backup path (backup path can repair link failures betweenvalue of the backup_count field is increasedintermediate node does not contain this bacthe intermediate node looks for another baneighbor table that can pass through anotherto the upstream node (Fig. 3(b)). If this baand can repair link failure 1 hop away, tbackup_count field is increased by 1, and node will not change the value of the baconly when it does not contain any bacdetermining the backup_count field in the Rrecording it in the routing table entry, the inupdates the reverse_border field of theupstream node and rebroadcasts this RREQ

When an intermediate node receives a packet, instead of discarding it immediconventional AODV, each copy is analyzwhether it provides more backup paths to than previous route. The hop count anpiggybacked on the RREQ are examined. equals the previous hop count, but backupthan the previous backup_count, the revforwards the RREQ will be adopted as route. The other duplicate RREQs are discthe generation of an additional routing loopto find the shortest reverse path containing paths to the source.

When a destination receives the first RREneighbor, the destination updates its routinggenerates a RREP packet. This RREP pacback to the source node to initialize daWhen the destination receives a duplicatfrom its neighbors, the new route is used backup_count - hop count, piggybacked ogreater than that of the previous route. (Theto select a route that has the shortest roumany backup paths to the source.) The desta RREP packet with a new sequence numbeRREP to the source. This work discardduplicate RREQs. Thus, a proper shorteslargest number of backup paths can be acqu

When an intermediate node receives a RRa neighbor, it adds a routing entry to its record the discovered route to the destinatentry sets the value of next hop field to the nthe RREP came, sets the border fieldpiggybacked on the RREP, updates the boRREP to the next hop field and forwmessage. As the RREP returns to the soalong the path not only contains a next hopthe downstream node but also contains ano

ld of the RREQ

ghbor table; any o connect to the (Fig. 3(a)). This n 2 hops, and the d by 2 unless the ckup path. Thus,

ackup path in the r node to connect ackup path exists the value of the the intermediate

ckup_count field ckup path. After RREQ packet and ntermediate node e RREQ to its packet. duplicate RREQ

iately as in the zed to determine

the source node nd backup_count

If the hop count p_count is larger verse route that the new reverse

carded to prevent p. The purpose is the most backup

EQ packet from a g table entry and cket is then sent ata transmission. te RREQ packet when the value,

on the RREQ is e value is defined uting length and tination generates er and replies the ds all the other st path with the

uired. REP packet from routing table to

tion. The routing node from which d to that field

order field of the wards the RREP ource, each node p field to indicate other field called

the border field to indicate the bnext hop field and the border fieldpath maintenance phase. If an intduplicate RREPs, it updates its rpropagates the RREP only whengreater sequence number than the way, intermediate node can supprereceived. This decreases the numbetoward the source node while also efor the quickest route is up-to-datthen begin data transmission as soreceived, and can update its routilearns of a route containing moprevious route.

C. Path maintenance Data packets are delivered via the

primary route is disconnected. Whfailure, it utilizes the backup pathneighbor table to repair the disruptecan connect to the border node infailure can be repaired by this bHowever, when the node does not cthe node looks for another backup the downstream node through anofailed link (Fig. 4(b)). Data padelivered through one or more badropped when route breaks occur.

Figure 2. (a) The format of revised RREQrevised RREP message. (c) The structure o

Figure 3. (a) Backup path discovery of nonode A. (b) Backup path discovery of node

border node. Both of the d are utilized during the ermediate node receives routing information and n the RREP contains a

previous RREP. In this ess all the other RREPs er of RREPs propagating ensuring that information te. The source node can oon as the first RREP is ing information when it ore backup paths than

e primary route unless the hen a node detects link

hs contained in its 2-hop d link. If the backup path n the routing entry, link backup path (Fig. 4(a)). contain this backup path, path that can connect to

other node to repair the ackets can therefore be ackup paths and are not

Q message. (b) The format of of revised routing table entry.

ode C with reverse border e Q with upstream node P.

256

The node that detected the link break butany backup path sends a RERR packet tinitiate a route re-discovery procedure. Tthese backup paths is that they are geograthe primary path, thereby providing efficienroute failures and maintaining an adequate ro

IV. SIMULATION In this section, we evaluate and compare

of the modified AODV scheme to the Aconventional AODV using NS-2 [8].

The schemes are evaluated using the performance metrics: 1) Packet delivery fraction － the number

packets divided by the number of generate2) Average end-to-end delay － the averag

all data packets to be transmitted across source to destination.

3) Control overhead － the number of transmitted per data packets delivered.

A. Simulation environment In our simulation, the MAC protocol is

802.11 and the propagation model is a reflection model. Table I lists the simulationsimulation scenario consisted of 100 mobiledistributed in a 1000m × 1000m square speed of the nodes is varied in five difmoving speeds: 0/5/10/15/20 m per secononly the continuous mobility case. The transeach node is 250m. There are 20 constantraffic sources distributed randomly over tCBR data packets are 512 bytes, and the packets per second. Simulations are run for 3

B. Simulation results This work evaluates the ability of AODV

the proposed scheme to change the topologychanging the node mobility through vmaximum speed.

Figure 4. (a) Link failure repair of node A with b(b) Link failure repair of node P with downstr

t does not contain to the source to

The advantage of aphically close to nt recovery from outing length.

the performance AOMDV and the

following three

of received data d data packets.

ge time taken for a network from

control packets

s based on IEEE two-ray ground

n parameters. The e nodes randomly

area. Maximum fferent maximum nd. We consider smission range of nt bit rate (CBR) the network. The sending rate is 4 300 seconds.

V, AOMDV and y dynamically by variation of the

TABLE I. SIMULATION

Simulator Simulation time Simulation area

Number of nodes Transmission range

Max speed CBR flows

Data payload Sending rate

Movement model

Although the performances of those in the static case, their differinmore apparent at higher speeds.

Figure 5 illustrates the comparisoin which AODV-BBS represents thebackup routing scheme. Simulation AODV-BBS and AOMDV outperfodoes not use a backup route; thus, da link is disrupted along the routeBBS utilize backup scheme to repathat data delivery ratio is increasedmobile cases, AODV-BBS loseAOMDV. The observation of a fraction in AODV-BBS implies thcontaining more available backup p

Figure 6 demonstrates the compaend delay where AODV incurs tthese protocols due to frequent dynamically changing topology. Sand AOMDV utilize backup padiscovery procedure whenever a linroute, the average end-to-end delathe average end-to-end delay of lower than that for AOMDV. Thisthat, in AODV-BBS, the backup pclose to the primary path and an admaintained. Consequently, the linkefficiently and the average end-towell.

Figure 7 compares the controresults indicate that when nodes control overhead of AODV-BBS that of AODV. An increasing mobioverhead in AODV-BBS and AOMowing to reduction of the route disroutes. Restated, in both AODV- Broute discovery is invoked onlyTherefore, AODV-BBS and AOMThe overhead of AODV-BBS is slAOMDV owing to that, in AODVmaintained by a neighbor table pe

border node C. ream node Q

N PARAMETERS

NS2 300 s

1000m x 1000m 100

250 m 0, 5, 10, 15, 20 m/s

20 512 bytes

4 packets/s Random waypoint

all protocols resemble ng performances become

on of data delivery ratio, e proposed AODV-based results indicate that both

orm AODV since AODV data loss occurs whenever e. AOMDV and AODV-ir the disrupted link such

d obviously. However, in es fewer packets than

higher packet delivery hat it can select a route aths than AOMDV does. arison of average end-to-he largest delay among route discoveries for a

Since both AODV-BBS th to reduce route re-nk is disrupted along the ay is reduced. However, AODV-BBS is slightly observation is owing to paths are geographically dequate routing length is k failure can be repaired o-end delay reduced as

ol overhead. Simulation stop or move slowly,

and AOMDV resembles le rate lowers the control

MDV than that in AODV scovery by using backup BS and AOMDV, a new

y when all paths fail. MDV outperform AODV.

ightly lower than that of V-BBS, backup paths are eriodically. Doing so can

257

avoid use of unavailable backup paths to initiate a new route discovery procedure and produce additional overhead.

V. CONCLUSION This work presents a novel AODV-based backup

routing scheme for MANETs. The proposed scheme utilizes 2-hop neighbor knowledge to establish backup paths during the route discovery phase and maintain updated knowledge of such links. These backup paths are geographically close to the primary path that can repair disrupted links locally without activating a route re-discovery procedure. Additionally, the proposed scheme selects the shortest path containing the largest number of backup paths to provide efficient recovery from route failure and maintain an adequate routing length. Simulation results indicate that the proposed backup routing scheme obtains a lower average end-to-end delay and less routing overhead than those of the AOMDV and the conventional AODV.

REFERENCES [1] C. E. Perkins and E. M. Royer, “Ad hoc on-demand distance vector

(AODV) routing”, in Proc. IEEE workshop on Mobile Computing Systems and Applications (WMCSA), Feb. 1999. pp. 90-100.

[2] C.E. Perkins and E.M. Royer, “RFC3561: Ad hoc on-demand distance vector (AODV) routing”, Internet RFCs, RFCs, Jul. 2003.

[3] E.M. Royer and C.K. Toh, “A review of current routing protocols for ad hoc mobile wireless networks”, in Proc. IEEE Wireless Communications, Apr. 1999, pp. 46-55.

[4] H. Liu and D. Raychaudhuri, “Label switched multi-path forwarding in wireless ad-hoc networks”, in Proc. IEEE International Conference on Pervasive Computing and Communications Workshops, Mar. 2005, pp. 248–252.

[5] Z. Ye, S. V. Krishnamurthy, and S. K. Tripathi, “A Framework for Reliable Routing in Mobile Ad Hoc Networks”, in Proc. IEEE INFOCOM, Mar. 2003, pp. 270 - 280.

[6] D. Ganesan, R. Govindan, S. Shenker, and D. Estrin, “Highly-resilient, energy-efficient multipath routing in wireless sensor networks”, in Proc. ACM SIGMOBILE Mobile Computing and Communications Review, Oct. 2001, pp. 11–25.

[7] M. K. Marina and S. R. Das, “On-demand multipath distance vector routing in ad hoc networks”, in Proc. International Conference for Network Procotols (ICNP), Nov. 2001, pp. 14–23.

[8] VINT Group, UCB/LBNL/VINT Network Simulator ns-2. [Online]. Available: http://www.isi.edu/nsnam/

[9] S. Motegi and H. Horiuchi, “AODV-based multipath routing protocol for mobile ad hoc networks”, in Proc. IEICE Transactions on Communications, Sep.2004, pp.2477–2483.

[10] J. Li, J. Jannotti, D. De Couto, D. R. Karger, and R. Morris, “A scalable location service for geographic ad hoc routing,” in Proc. ACM Int. Conf. Mobile Computing and Networking, Aug. 2000, pp. 120–130.

[11] A. A. Pirzada, C. McDonald, and A.Datta, “Performance Comparison of Trusted-Based Reactive Routing Protocols,” in Proc. IEEE Trans. On Mobile Computing, Jun. 2006, pp. 695–710.

[12] A. Valera, W. Seah and S.V. Rao, “Cooperative packet caching and shortest multipath routing in mobile ad hoc networks,” in Proc. IEEE INFOCOM, Apr. 2003, pp. 260–269.

60

70

80

90

100

0 5 10 15 20

AODVAOMDVAODV-BBS

Pack

etde

liver

y fr

acrio

n (%

)

Max. speed (m/s)

0

0.2

0.4

0.6

0.8

0 5 10 15 20

AODVAOMDVAODV-BBS

Max. speed (m/s)

Ave

rage

end

-to-e

nd d

elay

(sec

)

15

25

35

45

55

65

0 5 10 15 20

AODVAOMDVAODV-BBS

Max. speed (m/s)

Con

trol o

verh

ead

(pac

kets

)

Figure 5. Comparison of packet delivery fraction.

Figure 6. Comparison of average end-to-end delay.

Figure 7. Comparison of control overhead.

258