A Power-Aware Multicast Routing Protocol for Mobile Ad Hoc Networks With Mobility Prediction

Wireless Pers Commun (2007) 43:1479–1497DOI 10.1007/s11277-007-9321-0

A Power-Aware Multicast Routing Protocol for Mobile AdHoc Networks With Mobility Prediction

Nen-Chung Wang · Jong-Shin Chen ·Yung-Fa Huang · Yu-Li Su

Received: 29 July 2006 / Accepted: 1 May 2007 / Published online: 13 July 2007© Springer Science+Business Media, LLC 2007

Abstract A mobile ad hoc network (MANET) is adynamically reconfigurable wireless network that doesnot have a fixed infrastructure. Due to the high mobilityof nodes, the network topology of MANETs changesvery fast, making it more difficult to find the routes thatmessage packets use. Because mobile nodes have lim-ited battery power, it is therefore very important to useenergy in a MANET efficiently. In this paper, we pro-pose a power-aware multicast routing protocol (PMRP)with mobility prediction for MANETs. In order to se-lect a subset of paths that provide increased stability andreliability of routes, in routing discovery, each nodereceives the RREQ packet and uses the power-awaremetric to get in advance the power consumption oftransmitted data packets. If the node has enoughremaining power to transmit data packets, it uses the

N.-C. Wang (B)Department of Computer Science and InformationEngineering, National United University, Miao-Li 360,Taiwan, R.O.C.e-mail: [email protected]

Y.-F. Huang · J.-S. ChenGraduate Institute of Networking and CommunicationEngineering, Chaoyang University of Technology,Taichung 413, Taiwan, R.O.C.e-mail: [email protected]

Y.-L. SuDepartment of Computer Science and InformationEngineering, Chaoyang University of Technology,Taichung 413, Taiwan, R.O.C.e-mail: [email protected]

global positioning system (GPS) to get the locationinformation (i.e., position, velocity and direction) of themobile nodes and utilizes this information to calculatethe link expiration time (LET) between two connectedmobile nodes. During route discovery, each destinationnode selects the routing path with the smallest LET anduses this smallest link expiration time as the route expi-ration time (RET). Each destination node collects sev-eral feasible routes and then selects the path with thelongest RET value as the primary routing path. Then thesource node uses these routes between the source nodeand each destination node to create a multicast tree. Inthe multicast tree, the source node will be the root nodeand the destination nodes will be the leaf nodes. Sim-ulation results show that the proposed PMRP outper-forms MAODV (Royer, E. M. & Perkins, C. E. (1999).In Proceedings of the ACM MOBICOM, pp. 207–218,August 1999.) and RMAODV (Baolin, S. & Layuan,L. (2005). In Proceeding of the 2005 IEEE Interna-tional symposium on microwave antenna, propagationand EMC technologies for wireless communications,Vol. 2, pp. 1514–1517, August 2005.).

Keywords Global positioning system · Mobile adhoc networks · Power-aware routing · Multicastrouting · Tree-based routing

1 Introduction

A mobile ad hoc network (MANET) [20] is aself-organizing, dynamically reconfigurable wireless

123

1480 N.-C. Wang et al.

network that has no fixed infrastructure or centralmanagement. Two nodes communicate directly if theyare within transmission range of each other. Due tothe limited radio propagation range of wireless de-vices, routes are often “multi-hopped.” Every node in aMANET must be able to function as a router which canforward data packets to other nodes.

When applications must send the same data to morethan one destination, multicasting is often used. Mul-ticasting reduces the communication costs for appli-cations that send the same data to multiple recipients.Instead of sending data via multiple unicasts, multi-casting minimizes link bandwidth consumption, senderand router processing, and delivery delay. The stan-dard multicast routing protocols used in fixed networksor infra-structured mobile networks cannot be used inMANETs. Recently, many multicast routing protocolsfor MANETs have been proposed [2,4,8,10,12,14,17,18,21,24,29–31].

Multicast routing protocols in MANETs can be clas-sified into tree-based routing [2,8,11,12,17–19,21,24]and mesh-based routing [4,14,30]. Tree-based routingprotocols build a tree structure that connects all mul-ticast members and provide one path between a pairof source node and destination node. Examples of tree-based routing protocols include the robust multicastingusing an underlying link state [11], the multicast ad hocon demand distance vector routing protocol (MAODV)[24], the reliability of the multicast ad hoc on-demanddistance vector routing protocol (RMAODV) [2], therobust tree-based multicasting in ad hoc networks (RO-MANT) [29], the robust and cost-efficient group com-munication using overlay multicast in mobile ad hocnetworks (AOM) [12], supporting reliable multicastin mobile ad hoc networks [17], the dynamical con-struction of a core-based group-shared multicast treein mobile ad hoc networks [18], and the efficient rout-ing protocol for mobile ad hoc networks with neighborawareness and multicasting (NAMP) [21]. Mesh-basedprotocols build a mesh structure that connects the mo-bile nodes between the source node and the destinationnode to each other. Because the mesh-based multicastprotocols use redundant paths between two nodes, theyprovide alternative paths, and therefore a link failureneed not trigger the re-computation of a mesh, but in-stead increases network overhead by flooding througha mesh. Examples of mesh-based protocols include theon-demand multicast routing protocol (ODMRP) [14],

preemptive multicast routing (PMR) [30], the dynamiccore based multicast routing protocol for ad hoc wire-less networks (DCMP) [4].

In tree-based routing protocols, there is one pathbetween the source node and the destination node.However, tree-based routing protocols are not necessar-ily suitable for multicasting in MANETs, wherenetwork topology changes frequently. In such an envi-ronment, mesh-based routing protocols seem to outper-form tree-based routing protocols due to the availabilityof alternative paths which allow multicast datagrams tobe delivered to the receivers even if a link fails. On theother hand, mesh-based protocols have higher overheadthan tree-based protocols.

In MANETs, each mobile node has limited batterypower. In order to maximize the lifetime of ad hoc net-works, traffic should be sent via routes that can avoidnodes with low power while minimizing the total trans-mission power. Power consumption in a battery-pow-ered node generally falls into one of two categories:communication-related power and non-communication-related power. Non-communication-related power isvery dependent upon hardware implementation, rout-ing, and link protocol design. Examples of power-awarerouting protocols include minimum total transmissionpower (MTPR) routing [25,28], a power managed basedmulticast routing protocol [7], energy conserving mul-ticast [1], minimum drain rate (MDR) protocol [13],approximate minimum-energy multicasting [15], power-controlled hybrid multicast [3], constructingenergy-efficient multicast trees with delay constraints[20], online multicasting for network capacity maximi-zation [16], min-max battery cost routing (MMBCR)[26], and conditional max-min transmission batterycapacity routing (CMMBCR) [27].

In this paper, we propose a power-aware multicastrouting protocol (PMRP) with mobility prediction. Thisscheme takes into account the power parameter anddetermines the duration of time between two connectedmobile nodes using the global positioning system(GPS). It selects the routing paths with the longestduration of time for transmission to increase routingreliability. When a link on a routing path is broken, therouting path will be disconnected. We also propose aGPS-aided route reconstruction process that selects abackup path for route maintenance.

The rest of this paper is organized as follows. Wediscuss the related work in Sect. 2. Sect. 3 presents the

123

A Power-Aware Multicast Routing Protocol for Mobile Ad Hoc Networks With Mobility Prediction 1481

Fig. 1 Route discovery of the MAODV protocol

system model of this work. The proposed scheme ispresented in Sect. 4. Sect. 5 describes the simulationresults. Finally, Sect. 6 gives the conclusions.

2 Related work

Different from unicast communications, multicasting isthe networking technique of delivering the same packetsimultaneously to a group of destinations, it is calledselective broadcast. Multicasting reduces the commu-nication costs for applications that send the same datato multiple recipients. In stead of sending via multipleunicasts, multicasting minimizes the link bandwidthconsumption, sender and router processing, and deliv-ery delay. Due to the limits in MANETs, the traditionalwired multicast routing protocols cannot be used inMANETs.

2.1 Multicast Ad Hoc On-Demand distance VectorProtocol (MAODV)

In MAODV [24], the researchers directly followed theunicast AODV protocol [22] and discovered multicastroutes on demand using a broadcast route discoverymechanism that employs the same route request(RREQ) and route reply (RREP) packets that exist inthe unicast AODV protocol.

When a mobile node wishes to join a multicast tree orhas data to send to a multicast group but has no route itcan use, it will broadcast RREQ packet. Only a memberof the desired multicast group may respond to a RREQ.If an intermediate node receives a join RREQ for a mul-ticast group of which it is not a multicast member, or ifit does not have a route to the group, it will rebroadcastthe RREQ to its neighbors. If a node receives a RREQfor a multicast tree, it may reply if it is a member ofthe multicast tree and its recorded sequence numberfor the multicast group is as great as that contained inthe RREQ. The responding node of the multicast tree

unicasts a RREP back to the source node after it re-ceives a RREQ packet. As nodes along the path to thesource receive the RREP, they add both a route table anda multicast route table entry for the node from whichthey received the RREP, as shown in Fig. 1.

After the source node broadcasts a RREQ packetto a multicast group, it often receives more than oneRREP packet. For a period of time, the source nodekeeps the received route with the greatest sequencenumber and shortest hop count to the nearest multicastmember of the multicast tree, and disregards the otherroutes. At the end of this period, it enables the selectednext hop node in its multicast route table, and sends aunicast activation message (MACT) to the selected nexthop node. On receiving the message, the next hop nodeenables the entry for the source node in its multicastrouting table. The next hop node does not propagatethe message any further if it is a member of the mul-ticast tree. However, if the intermediate node is nota member of the multicast tree, it will have receivedseveral RREPs from its neighbors. It keeps the bestnext hop node for its route to the multicast group, uni-casts MACT to that next hop node, and enables thecorresponding entry in its multicast routing table. Thisprocess continues until the node that originated thechosen RREP is reached. MACT ensures that the mul-ticast tree does not have multiple paths to any tree node.The intermediate node forwards data packets only alongthe activated routes.

2.2 On-Demand Multicast Routing Protocol(ODMRP)

The ODMRP [14] is mesh-based routing protocol forMANETs, it establishes on-demand multicast route anduses a forwarding group concept. Data is forwardedby using restricted flooding, as show in Fig. 2. Mul-ticast members to create routing information initiatethe request phase. Sender floods a join query packet toentire network to refresh membership. When an inter-mediate node receives a join query packet which it isnot a receiver, it will stores the backward learning intorouting table and rebroadcasts the packet. Finally whenjoin query packet reaches a receiver, the receiver cre-ates and broadcasts a join reply packet to its neigh-bors. When a node receives the join reply packet, itchecks whether the next node id in join reply packetmatches it own. If yes, it sets the forwarding group flag,

123


Fig. 2 Mesh creation inODMRP. (a) Join-Querypacket propagation. (b)Multicast mesh

it will become a part of the forwarding group. Join replypacket thus propagated by forwarding group memberuntil it reaches source via a shortest path.

2.3 The Reliability of the Multicast Ad HocOn-Demand Distance Vector Routing Protocol(RMAODV)

The (MAODV) routing protocol [24], designed formobile ad hoc networks, offers quick adaptation todynamic link conditions, low processing and memoryoverhead, and low network utilization. The reliabil-ity of the multicast ad hoc on-demand distance vec-tor (RMAODV) routing protocol [2] in terms of thedelivery of data packets. RMAODV supports a reli-able multicasting suitable for mobile ad hoc networkby reducing the number of route reconstructions andpacket retransmissions.

Protocol relays are placed along the path. Each relaynode stores the forwarded packets in its memory for a fi-nite time. An extra sequence number is assigned to eachpacket for protocol relay. The protocol relay analyzesthis extra sequence number on every packet that propa-gates through it. When a loss is detected by a gap in thesequence, the relay node immediately requests retrans-mission of the missing packets from the previous relaynode. It uses a NACK to request the retransmission.

In Table 1, we give the comparisons of some famousmulticast protocols, such as MAODV, ODMRP, andRMAODV.

3 System Model

In this section, we first describe the technology used inthe GPS and the mobility prediction mechanisms. Then

we describe the power-aware metric used to computesthe power consumption. Finally, we present the mul-ticast ad hoc on demand distance vector (MAODV)routing protocol.

3.1 Global Positioning System (GPS)

The Global Positioning System (GPS) [5,11] is verypopular for location determination of a specific target,recently. If we have a mobile device, like a PDA or anotebook, that has a GPS receiver, we can know ourlongitude and latitude. By matching the known longi-tude and latitude to the map, we can obtain the position.By continuously updating your position, a GPS receivercan also provide data on your speed and direction oftravel.

GPS is the only system available today that is able toshow the exact position of an object or person anywhereon earth at anytime in any weather. It is a satellite-based,radio navigation system. The satellites are continuouslymonitored by ground stations located worldwide. Thesatellites transmit signals that can be detected by any-one with a GPS receiver. Using a receiver, one candetermine the location of an object or person with greatprecision.

GPS consists of three segments: the space segment,the user segment, and the control segment. The spacesegment consists of 24 satellites, each in its own or-bit 11,000 nautical miles above the Earth. The usersegment consists of receivers which one can hold inone’s hand or mount in one’s car. The control segmentconsists of ground stations (five ground stations, lo-cated around the world) that make sure the satellites areworking properly.

123


Fig. 3 Power first orderradio model

TransmitElectronics

Tx Amplifer

ETx(d)

Eelec × k Eamp × k × d2

ReceiveElectronics

ERx

Eelec × kd

k bits packet k bits packet

Table 1 The comparisons of multicast protocols

Protocol MAODV ODMRP RMAODV PMRP

Category Reactive Reactive Reactive ReactiveConfiguration Tree Mesh Tree TreeMaintain mode Destination-initiated Source-initiated Destination-initiated Destination-initiatedUnicast dependency Yes (AODV) No Yes (AODV) Yes (AODV)Periodically update Yes Yes Yes YesControl packet flood Yes Yes Yes YesReliability No No Yes YesMobility prediction No No No YesPower-aware No No No Yes

Table 2 Parameters used in the simulations

Parameter Values

Examined protocols PMRP, MAODV, RMAODVSimulation area 1000 m×1000 mMulticast group size 5–40Mobility speed 0–20 m/sMobility model Random waypoint modelPropagation model Free space propagation modelNode transmission range 250 mData packet size 512 bytes

GPS receivers typically work well outdoors and canprovide positioning accuracy within a 15-meter range.Assistance from ground stations can improve accuracy.Such systems, called differential GPSs, can reduce theerror to less than a few meters.

The GPS system has satisfied performance in manylocation aware applications. It usually used in a com-plex environment, include out door and indoor. For out-door applications, Global Positioning System (GPS)has proven reliable and accurate. For urban and indoorapplications various RF based ranging/positioning tech-niques have been investigated. Recently modificationsto GPS for indoor applications have been proposed.For example, network assisted GPS (A-GPS) [1,6] pro-vides via a cellular data link, additional ephemeris tostand alone GPS for aiding. It is accurate within 50meters when users are indoors and 15 meters when theyare outdoors, well within federal guidelines and an

order of magnitude more sensitive than conventionalGPS. Increase in the indoor GPS receiver sensitivityhas been proposed by providing a large number of cor-relators. A number of other sensors such as InertialMeasurement Unit (IMU), enhanced dead reckoningdevices, miniature or micro electromechanical systems(MEMS) could come to aid for durations following theloss of GPS. In addition path constraints based on theindoor environment may assist in determining accu-rately the position.

3.2 Mobility Prediction Mechanisms

In this section, we introduce the mobility predictionmethod. This method uses the location informationprovided by GPS. We assume a free space propaga-tion model [23] in which the signal strength dependssolely on the distance to the transmitter. The free spacepropagation model assumes the ideal propagationcondition that there is only one clear line-of-sight pathbetween the transmitter and receiver. The free spacemodel basically represents the communication rangeas a circle around the transmitter. If a receiver is withinthe circle, it receives all packets. Otherwise, it losesall packets. We also assume that all nodes have theirclocks synchronized using the GPS clock. If we knowthe motion parameters of two nodes, we can calculatethe duration of time these two nodes remain connected.The speed and heading of a mobile can be obtained from

123


the mobile own instruments and sensors (e.g., compass,odometer, speed sensors).

We assume that two nodes A and B are within thesame transmission range r of each other. We let (x1,y1) be the coordinate for mobile node Aand (x2, y2) bethe coordinate for mobile node B. We let v1 and v2 bethe mobility speeds and θ1 and θ2 (0 ≤ θ1, θ2 < 2π)

be the moving directions for mobile node A and B,respectively. We can obtain the duration of time Dt byusing the following equation [23]:

Dt = −(ab + cd) + √(a2 + c2)r2 − (ad − cb)2

(a2 + c2)

(1)

Note that a = v1 cos θ1−v2 cos θ2, b = x1−x2, c =v1 sin θ1 − v2 sin θ2, and d = y1 − y2. In addition, theequation cannot be applied when v1 = v2 and θ1 = θ2,and when Dt is ∞. In order for the information fromthe GPS to be utilized, the packets must include extrafields. When a source node sends a request packet, thepacket appends its location, direction, and speed. Thenext hop of the source node receives the request packetto predict the link expiration time between itself andthe source node. If node Bis the next hop of the packetfor node A, node A will insert its location informationin the packet so node B will be able to compute the linkexpiration time between node A and node B.

3.3 Power-Aware Metric

In a MANET, the power consumption of each mobilenode becomes an important factor that affects networkperformance. A lack of sufficient hosts can result in apartition of the network, causing interruptions in com-munications. In a simple radio model [9], Eelec = 50nJoule/bit is dissipated to run the transmitter or receivercircuitry. Eelec is the power consumption of the circuititself. Assuming d2 energy loss, where d is the distancebetween nodes, a transmission amplifier at the sendernode further consumes further Eampd2, where Eamp =100 pJoule/bit/m2. Eamp is the power consumed by theamplifier to transmitting packets. These parameters areslightly better than the parameters used in current state-of-the-art in radio design. We also assume an r2energyloss due to channel transmission. Thus, to transmit ak-bit message a distance d using the radio model, theradio expends:

ET x(k, d) = Eelec × k + Eamp × k × d2 (2)

and to receive this message, the radio expends:

ERx(k) = Eelec × k (3)

Receiving a message is not a low cost operation usingthese parameter values. Protocols should thus try tominimize not only the transmission distances but alsothe numbers of transmission and reception operationsfor each message. We can generalize about the totaltransmission consumption as follows:

Etotal(k) = Eelec × k + Eamp × k × d2

+ (Eelec × k) (4)

Figure 3 shows the power-aware metric of a firstorder radio model.

4 Power-Aware Multicast Routing Protocol(PMRP) With Mobility Prediction

In this section, we propose a power-aware multicastrouting protocol (PMRP) with mobility prediction. Theproposed protocol includes the route discovery processand the route maintenance process.

4.1 Tables for Routing

In PMRP, we use three kinds of tables in our routingprotocol: the neighboring node table, the current pathtable, and the maintenance table. The details of thesetables are described below.

(a) Neighboring node table: A node will record theinformation of other nodes that are within thetransmission range if it can hear the “beaconpacket.” The format of the table is <node_ID, thelocation information of node_ID, distance>.

(b) Current path table: This table contains the cur-rent path that is used for the transmission of data.The format of the current path table is <seq_num-ber, path_table, route expiration time>. Here, theformat of the path_table is <node_IDi , Dti,i−1>.Node_IDi is a node number in the MANET andDti,i−1 is the duration of time between node_IDi

and node_IDi−1 on the routing path.(c) Maintenance table: This table contains the loca-

tion information that is used in our maintenanceprocess. The format of the maintenance table is

123


Destination node

Mobile node

Source node

RREQ

RREP

Fig. 4 On-demand process for multicast group management

<Source ID, Destination ID, location informationof the last node, location information of the nextnode>. We can update the entry by sending a datapacket.

4.2 Route Discovery Process

In PMRP, as shown in Fig. 4, the member nodes andmulticast routes are established and updated by thesource “on demand.” Similar to on-demand unicastrouting protocols, PMRP contains the request phaseand reply phase. Since PMRP is an on-demand proto-col, it sets up routes when a source has data to send.

First, we define the format of the RREQ, RREP,and beacon packets, as shown in Figs. 5–7. The RREQpacket is broadcast by the source node when the source

node needs to send data to the destination nodes andhas no route information in its routing table. When eachdestination node can finds a suitable path, the destina-tion node sends a RREP packet to the source node. Thebeacon packet is broadcast by each node frequentlyto determine the neighboring nodes and the distancebetween that node and the neighboring nodes.

RMRP’s RREQ is the same as standard RREQ ex-cept that RMRP’s RREQ have three addition fields:location information, data information, and link expi-ration time. RMRP’s RREP is the same as standardRREP except that RMRP’s RREP have two additionfields: link expiration time and route expiration time.

In our proposed PMRP, we define three parameters:the total power consumption of transmitted data pack-ets (Pprediction), the link expiration time (LET) andthe route expiration time (RET). The Pprediction repre-sents the total power consumption of transmitted datapackets. It is calculated by each node, when a nodereceives a RREQ packet using Eq. 4. The LET repre-sents the link duration time between two nodes. Whena node forwards a RREQ packet, an intermediate nodereceives this packet and calculates the LET by usingEq. 1. The RET is equal to the minimum of the setof LETs for the routing path. Thus, the RET is thetime that the route is expected to be stable. In general,the route with larger RET is the more reliable routingpath.

In PMRP, each node broadcasts a beacon packet tofind out the neighboring nodes, when a neighboring

Fig. 5 RREQ packet format.

123


Fig. 6 RREP packet format

Fig. 7 Beacon packet format

node receives a beacon packet. Periodic beacon pack-ets are used by the mobile node to obtain the neighbor-hood information. In this manner, the beacon packetto find out the neighboring nodes, when a neighboringnode receives a beacon packet, it calculates the dis-tance between the two nodes. We call the period forsend the beacon packets. It must be long so that it doesnot induce much overhead on the network.

When a source node wants to multicast informa-tion to some multicast destination nodes, in the begin-ning, the source node will be the root node of themulticast tree and will initiate a path discovery processby broadcasting a RREQ packet to its neighbors. TheRREQ packet has to include the location information,that is, the source node’s location, velocity, and direc-tion. Once an intermediate node receive the RREQ,it first checks the broadcast ID to determine whetherthe entry is its own ID. If not, it will calculate thePprediction, the power consumed when a packet is trans-mitted to its neighboring nodes. If the power remainingon the intermediate node is lower than the Pprediction

(Premain < Pprediction), it means that the intermediatenode does not have enough power to transmit data pack-

ets. It will not rebroadcast the RREQ packet and willdiscard the packet. If the power remaining on the nodeis higher than the Pprediction (Premain > Pprediction),the node will calculate the LET between the next-hopnode using the location information and then write thedata into the LET field of the RREQ packet and itsrouting table. It will then rebroadcast the packet to itsneighbors. By the time a request packet arrives at thedestination node, it has recorded all the nodes along therouting path it has traversed and the duration of time ofeach link along the route. Then the destination node willdetermine the RET by looking in the packet to find theminimum number of LETs along the path of the routerequest packet. Each destination node waits a periodof time to collect RREQs. We assume that each desti-nation node has many routing paths that can be used.Each destination node then selects the primary routingpath with the maximum RET and sends a RREP packetback to the multicast source node along the decidedpath. The intermediate nodes between each source anddestination pair may need to act as routers for the tree.We employ the routing paths with the longest routeduration of time to establish a reliable multicast tree.

123


The detailed route discovery process is shown inAlgorithm 1, and the steps of the route discovery pro-cess are described below.

Step 1: In the beginning, the source node broadcastsa RREQ packet to the neighboring nodes thatare in its transmission range. After the neigh-boring nodes receive the packet, each neigh-boring node calculates the total powerconsumption of transmitting data packets(Pprediction) to its neighboring nodes. If thepower remaining (Premain) on the node islower than the Pprediction, it will discard theRREQ packet.

Step 2: If the power remaining on the node is higherthan the Pprediction, the node will use thelocation information to calculate the linkexpiration time. Then the node adds its IDand the LET for the last link of the requestpacket to the packet entry, and rebroadcaststhe RREQ packet.

Step 3: When each destination node receives arequest packet, it waits for a period of time toreceive other request packets and then deter-mines the RET of the routing path by us-ing the minimum of the set of LETs alongthe route. Then each destination node selectsthe path with the maximum RET to be theprimary routing path, and sends a RREPpacket to the source node along the primaryrouting path.

Let us consider the example shown in Fig. 8. In Fig. 8(a),each node in the network broadcasts a beacon packetto obtain the information of the neighboring nodes’information (i.e. the neighboring nodes’ ID and dis-tance). Source node Awants to send data to destinationnodes J ,K ,L,M , andN . It broadcasts a RREQ packetto its neighboring nodes. Nodes B,C, …, H , and I

will append their own information, such as their ownID and the duration of time, to the request packet andthen forward the request packet. In this example, whensource node A broadcasts the RREQ packet, interme-diate node B receives the request packet and calculatesPprediction. Pprediction of the node B is 30 and Premain

of the node B is 50. Because Premain of the node B

is higher than Pprediction of the node B, node B hasenough power to transmit data packets. Then it willutilize the location information to get the LET betweennodes A and B, write the information into the RREQ

Algorithm 1: Route Discovery ProcessSuppose n is the number of mobile nodes and N is the set ofmobile nodes N = {N1, N2, . . . , Nn}. Assume that source nodeNi wants to find a path to destination node Nj . Node Ni broad-casts a route request packet, and node Nk receives the RREQpacket, let Ni,Nj,Nk ∈ N , 1 ≤ i, j, k ≤ n and i �= j .if (Node Nk is the destination node Nj )

{

(1) Node Nk counts the waiting time required to receive aRREQ packet.

(2) NodeNk checks the entry of the LET of the RREQ packetto find the RET of the path.

(3) Node Nk selects the path with the maximum RET as theprimary path and unicasts a RREP packet to the sourcenode. If two paths have the same RET, both of themselect the lower hop count (HC) to relay the packet.

(4) Each node receives the relay packet and writes the entryto the current path table.

(5) Node Ni starts to send data.

}else{

(1) Node Nk calculates the total power consumption oftransmitted data packets (Pprediction). If the powerremaining on the node is lower than the Pprediction, thenode will discard the RREQ packet.

(2) Node Nk calculates the duration of time and records thenode_ID and the LET to the RREQ packet.

(3) Node Nk forwards the RREQ packet to the neighboringnodes.

}

packet and then rebroadcast this packet. Pprediction ofnodes D and H are higher than Premain of nodes D

and H , respectively. It means that the nodes do nothave enough power to transmit data packets, and theywill discard these RREQ packets. Finally, destinationnode Kreceives three request packets. The first packetcontains path (A, C, F , K) with the duration of timeof LETs = (6, 4, 3). The second packet contains path(A, C, E, K) with LETs = (6, 7, 5). The third packetcontains the path (A, B, E, K) with LETs = (6, 4, 5).Node K can obtain the RET evoked from the mini-mum LET. In this case, the RET of path (A, C, F ,K) is 3, the RET of path (A, C, E, K) is 5, andthe RET of path (A, B, E, K) is 4. Thus, the RETof path (A, C, E, K) is larger than that of path (A,C, F , K) and that of path (A, B, E, K). Thus, path(A, C, E, K) is more stable than the other two paths.Similarly, nodes J ,L,M , andN will select paths (A,C,E, J ), (A, G, L), (A, G, M), and (A, G, I , N)

with the RETs equal to 5, 5, 5, and 4, respectively. If

123


Fig. 8 An example ofmulticast routing (a) theroute discovery process, (b)the route relay process, and(c) The reliable multicasttree

(a) (b)

Source node

Destination node

Mobile node

Route request (RREQ)

Route reply (RREP)

(c)

J

65

5

7

66

47

53

4

F

K

N

L

M

I

HG

A

B C

ED

4

55

6

J

F

K

N

L

M

I

HG

A

B C

ED

J

F

K

N

L

M

I

HG

A

B C

ED

{30,50}

{40, 35}

{35, 40}

{20, 60}

{25, 40}

{25, 35} {36, 30}

{40, 60}

5 5 3

47

66

5 6

7

5

4

56

7 4

7

{x, y} x: Pprediction y: Premain

{11, 45}

{11, 40} {11, 36}

{11, 60}{11, 30}

Premain of the destination node is lower than Pprediction

of the destination node, the route discovery processwill fail.

In Fig. 8b, each destination node selects the pri-mary routing path with the longest duration of timefor multicast routing and then sends its RREP packetback to the source node. Source node A receives all thedestination nodes’ reply packets and delivers the datapacket along these routing paths. Finally, as shown inFig. 8c, we can use these routing paths to build a multi-cast tree, the multicast tree not only has enough powerto transmit data packets but also has the longest routelifetime.

4.3 Route Maintenance Process

In the following, we present the route maintenanceprocess that includes multicast join operation, nodepruning operation, and broken link maintenance.

4.3.1 Multicast Join Operation

When a node wants to join a multicast tree, it broad-casts a join request packet across the networks. Only anode that is a member of a multicast tree (i.e., a routerfor the group) may respond. If a node receives a joinrequest packet for a multicast group of which it is not

123


Fig. 9 Multicast joinoperation (a) Join requestpacket propagation, (b) joinreply sent back to source,and (c) multicast tree branchaddition

a member, or if it receives a join request packet anddoes not have a route to that group, it creates a reverseroute entry to the prospective node and then broad-casts the join request packet to its neighbors. Eachmember node of the multicast tree waits for a periodtime to collect the join request packet, chooses the lon-gest duration of time route from the prospective node,and sends back the join reply packet. The join replypacket must add the LET information between the pro-spective node and the source node. The prospectivenode will select the longest duration of time route tojoin the multicast tree. Figure 9 shows the node joinoperation.

In Fig. 9a, prospective node L broadcasts a join re-quest packet to its neighbors. When nodes D, F , and G

receive the packet, they will calculate the LET with theprospective node, put the information into the packet,and then rebroadcast it, repeating this process until themulticast group member receives the packet. The mem-ber nodes collect these join request packets. Nodes A,B, C, E, and Iwill select the route with the longest linkexpiration time and return the join reply packet to theprospective node, as shown in Fig. 9b. The node waitsfor a period of time to collect these join reply packets

and then chooses the most reliable route to the tree, asshown in Fig. 9c.

4.3.2 Node Pruning Operation

When a node is removed from a multicast tree, thepruned node sends to its upstream node a quit_requestpacket. When the upstream node receives the quit re-quest packet, it will remove the corresponding entryfrom its multicast routing table. If the upstream nodebecomes a leaf node (because it is a router for the treeand not the tree receiver as a result of this removal, andbecause it is not interested in multicast traffic from thistree), it can further prune itself from the tree and in turnsend a quit request packet to its upstream node.

For example, as shown in Fig. 10a, node I decides toleave the multicast tree. It sends a quit request packet tonode C. When node C receives the packet and deletesnode I from its list of next hops, it discovers that it is aleaf node. But because it is just a router for the multicasttree and not a multicast member, it will in turn send aquit request packet to node A. Fig. 10b illustrates thenew multicast tree.

123


Fig. 10 Node pruningoperation (a) quit requestpacket propagation and (b)multicast tree after pruning

J

K

I

C

H

A

(a)

L

J

K

H

A

(b)

L

Source node

Destination node

Mobile node

Quit request

Multicast tree link

4.3.3 Broken Link Maintenance

In a MANET, because of the mobility of the mobilenodes, a link between two nodes will break easily. InPMRP, we use the location information to get the LETsand RETs. We can predict the time that the link willbreak. The maintenance process will be started beforethe link breaks. The upstream node of the link willbroadcast the route request_repair (RREQ_R) packetto the downstream node. The RREQ_R packet mustadd information about the remnants of the data packetsthat need to be sent. An intermediate node receives thepacket, and uses the remnant data information to calcu-late the Pprediction, if Premain is lower than Pprediction,it drops this RREQ_R packet. If Premain is higher thanPprediction, the node calculates the LET between thenext-hop node using the location information, writesthe LET into the link expiration time field of theRREQ_R packet, and rebroadcasts the RREQ_R packetto its neighbors until the request packet arrives at thedownstream node. When the downstream node receivesthe RREQ_R packet, it determines the RET by look-ing in the packet to find the minimum number of LETsalong the path of the RREQ_R packet. The downstreamnode waits a period of time to collect more RREQ_Rsand selects the alternative routing path with the max-imum RET. Then it sends the route reply_repair(RREP_R) packet back to the upstream node. Theupstream node will use the alternative path to trans-mit data before the original link breaks. Moreover, the

upstream node defines a threshold Ttimeout . Theupstream node will drop the RREP_R packets afterthe threshold Ttimeoutcounts down to 0. Then it sendsan error packet to the source node and restarts theroute discovery process. The formats of the RREQ_Rpacket and RREP_R packet are shown in Figs. 11and 12.

The detailed route maintenance process is shown inAlgorithm 2, and the steps of the route maintenanceprocess are described below.

Step 1: If a link is broken, the route maintenance pro-cess is executed to avoid the loss of a message.Each of the nodes will construct a mainte-nance table. A node can know the informationof the last node and the next node by broad-casting packets. When a node knows that thelink will break, it uses its own informationand the information of the next node to starta maintenance process.

Step 2: A node broadcasts a RREQ_R packet to thenext node. When an intermediate node receivesthe packet, it will calculate the Pprediction. IfPremain is lower than Pprediction, it will dropthis packet.

Step 3: If the power remaining on the node is higherthan the Pprediction, the node will use thelocation information to calculate the LET.Then the node adds its ID and the LET forthe last link of the request packet to the

123


Type

0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 10 1 2 3

1

2

3

4

Reserved Hop count

Break node address

Next break node address

Break node ID

5

6

7

8

9

Next break node ID

Broadcast ID

Remnant data information

Routing path

Link expiration time

Fig. 11 RREQ_R packet format

Fig. 12 RREP_R packet format

packet entry, and rebroadcasts the RREQ_Rpacket.

Step 4: When the next node receives the RREQ_Rpacket, it waits for a period of time to receivemore RREQ_R packets and determines theRET of the routing path by using the mini-mum of the set of LETs along the route. Thenthe next node selects the path with the max-imum RET to be the alternative routing pathand sends the route RREP_R packet to thefirst node. Until the routing path is broken,it sends data along the alternative path. If itdoes not find a replacement node, it sends anerror packet to the source node and restartsthe route discovery process.

An example of the broken link maintenance process isshown in Fig. 13a. Source node A sends a data packet tothe destination nodes H ,I ,J ,K , and L. We assume thatthe link between node B and node E will break. When

node B detects that the link between node Band nodeE will break, node B broadcasts a RREQ_R packet todiscover a backup path. In this example, node B broad-casts a RREQ_R packet to discover the backup routingpath. When node D and node J receive the RREQ_Rpackets, they update their routing tables and utilize thepower first order radio model to get Pprediction. Be-cause Premain of node J is lower than Pprediction, nodeJwill discard the RREQ_R packet. Because Premain ofnode D is higher than Pprediction, nodeD will use thelocation information to calculate the LET and write theLET into the link expiration time filed of the RREQ_Rpacket, and then rebroadcast the RREQ_R packet. Weassume the RREQ_R packet arriving at node E is vianode D. Node E selects the RET by looking in theRREQ_R packet to find the minimum number of LETs.Finally, node E sends back a RREP_R to node B. Whenthe link between node B and node E is broken, nodeB delivers data along the backup path (B,D, E) to thedestination nodes K andL, as shown in Fig. 13b. If we

123


Fig. 13 Repair of a brokentree link (a) link break and(b) repaired multicast tree

H

L

G

K

D

E

J

B

I

C

A

{20, 15}

{13, 30} H

L

G

K

D

E

J

B

I

C

A

Source node

Destination node

Mobile node

Multicast link

RREQ_R

RREP_R

{x, y} x: Pprediction y: Premain

(a) (b)

cannot find a replacement node, we use the last nodeof the broken path to send the error packet back to thesource node. We utilize the error packet to restart theroute discovery.

Algorithm 2: Route Maintenance ProcessSuppose that n is the number of mobile nodes and N is the setof mobile nodes N = {N1, N2, . . . , Nn}. Assume that node Ni

wants to send a packet to node Nj , where Ni,Nj,∈ N , 1 ≤i, j ≤ n, i �= j , and that the link between node Ni and nodeNj

will break.if (the link of node Ni to node Nj will break){

(1) Node Ni saves the current data packet.(2) NodeNi broadcasts the RREQ_R packet to node Nj ,

counts down Ttimeout seconds, and waits for theRREP_R packet to return.

if (the RREP_R packet is back in Ttimeout seconds){

(1) NodeNi uses the replacement path to replace thepath that will break.

(2) NodeNi continues packet transmission.

}elseNodeNi sends an error packet to the source and restarts the

routing discovery process.

}

5 Simulation Results

In this section, we first give some parameters used inour simulation. Then, we present the simulation resultsof the proposed scheme. We compare the performanceof the proposed PMRP with that of the MAODV [24].

5.1 The Simulation Environment

As shown in Table 2, we designed and implementeda simulator to act as an simulation platform for test-ing multicast operations in a MANET. The simula-tion modeled a network in a 1000 m×1000 m area with50 mobile nodes. A random waypoint model was usedin the simulation. The mobile speed of each node wasfrom 0 m/s to 20 m/s. The transmission range was 250 m.The data packet size was 512 bytes. The data transmis-sion rate was set to 2 Mbps. Each simulation was con-ducted for 600 s. The source and destination nodes wererandomly chosen. Each node was randomly assigned aninitial energy.

The performance evaluation metrics used in simula-tions were as follows:

1. Packet delivery ratio: The data packets delivereddivided by the data packets expected to be deliv-ered.

2. Control overhead: The control packets transmitteddivided by the data packets delivered.

3. Total overhead: The total packets transmitteddivided by the data packets delivered. The totalpackets is the control packets plus the data packets.

5.2 Performance Analysis

In the following, we will study the impact of mobilityspeed and multicast group size on MAODV, RMAODVand the proposed PMRP. Some simulations are con-ducted for packet delivery ratio, control overhead, andtotal overhead, respectively.

123


0.8

0.85

0.9

0.95

1

0

Mobility speed (m/sec)

oitaR yrevile

D tekcaP

PMRP

MAODV

RMAODV

2018161412108642

Fig. 14 Packet delivery ratio versus mobility speed

5.2.1 Impact of Mobility Speed

In this simulation, the multicast group size was as-sumed to be 20. The mobility speed varies from 0 m/sto 20 m/s. 5 sources transmit packets and total networktraffic load is fixed at 2 pkts/s.

Figure 14 depicts the packet delivery ratio of thesethree protocols under different mobility speeds. Thepacket delivery ratio decreased with increasing mobil-ity due to more link breaks. This resulted in more multi-cast tree partitions for PMRP, MAODV, and RMAODV.Notice that the packet delivery ratio of RMAODV washigh when the nodes had low mobility. This is becausethat the multicast tree structure has mostly static andthe protocol replay method been used. Therefore, thenumber of packet deliveries was high. At high speeds,the tree links broke down quite often, leading to con-stant branch reconstructions and larger packet losses.In PMRP, we used the power information and loca-tion information to select the more stable routing pathsfor multicasting. Thus, the packet delivery ratio of theproposed PMRP was higher than that of MAODV andRMAODV.

Figure 15 shows the control overhead under differ-ent mobility speeds for these three protocols. As wasexpected, the control overhead increased as the mo-bile speed increased. The reason is that there weremore chances for routes to break when the speed of themobile nodes was faster. Thus, the number of rebroad-casts increased. Because the proposed PMRP elimi-nates inefficient nodes in order to decrease the num-ber of control packets to be broadcast and selects themore stable route for data transmission, the number ofroute reconstructions was less. Therefore, the proposed

0

0.5

1

1.5

2

0 2 4 6 8 10 12 14 16 18 20


tekcaP derevileD lato

T / tekcaP lortnoC

o

PMRP

MAODV

RMAODV

Fig. 15 Control overhead versus mobility speed

0

1

2

3

4

0 2 4 6 8 10 12 14 16 18 20



T / tekcaP ataD

o

PMRP

MAODV

RMAODV

Fig. 16 Total overhead versus mobility speed

PMRP had have a lower control overhead than MAO-DV and RMAODV. In additions, the proposed PMRPcan maintain routing paths in advance by using the pro-posed route maintenance process.

Figure 16 shows the total overhead under differentmobility speeds for these three protocols. Both the pro-tocols show similar trends with the increased in mobil-ity speeds. The total overhead of the proposed PMRPwas lower than that of MAODV and RMAODV evenwhen the mobility speed increased. This is becausethe proposed PMRP utilizes the power-aware metricand mobility prediction to get the more reliable routingpaths.

5.2.2 Impact of Multicast Group Size

In this study, the mobility speed was assumed to be10 m/s. The multicast group size varies from 5 nodes to40 nodes. 5 sources transmit packets and total networktraffic load is fixed at 2 pkts/s.

123


0.8

0.85

0.9

0.95

1

5 10 15 20 25 30 35 40

Multicast group size

oitaR yr evile

D tekcaPo

PMRP

MAODV

RMAODV

Fig. 17 Packet delivery ratio versus multicast group size

0

0.5

1

1.5

2

5 10 15 20 25 30 35 40



T / tekcaP lortnoC

o

PMRP

MAODV

RMAODV

Fig. 18 Control overhead versus multicast group size

Figure 17 illustrates the packet delivery ratio underdifferent multicast group size for these three protocols.In this figure, as the multicast group size increased, thepacket delivery ratio slightly decreased. But, the groupsize does not have great impact on the packet deliveryratio. The packet delivery ratio of the proposed PMRPoutperforms that of MAODV and RMAODV.

Figure 18 shows the control overhead under differ-ent multicast group size for these three protocols. Wecan observe that the control overhead decreased whenthe multicast group size increased. The routing over-head of MAODV and RMAODV is higher than that ofthe proposed PMRP.

Figure 19 shows the total overhead under differentmulticast group size for these three protocols. In thisfigure, the total overhead decreased when the multicastgroup size increased. These the protocols show similartrends with the increased in multicast group size butthe proposed PMRP perform slightly better.

0

1

2

3

4

5 10 15 20 25 30 35 40



T / tekcaP ataD

o

PMRP

MAODV

RMAODV

Fig. 19 Total overhead versus multicast group size

0

20

40

60

80

100

120

140

160

0.25 0.5 0.75 1 1.25 1.5 1.75 2

Waiting period time (Sec)

stekcap tseuqer fo rebmu

N

PMRP

Fig. 20 Number of request packets versus waiting period time

Figure 20 shows the number of request packets ofPMRP with different waiting period time. The wait-ing period time is the time that the destination nodewaits for the receipt of the request packets after thefirst request packet arrive the destination node. In themulticast join operation, the member node of the mul-ticast tree waits for a period time to collect the joinrequest packet. If the period time is long, we will havemore paths for select but long delay. In another word,if the period time is short, we will have fewer paths forselect but faster. Therefore, in the simulation, we sim-ulate this issue. In result, we obtain the 1.0 s will havethe best effect.

In MANETs, higher mobility speed causes more linkbreakages, thus increasing control overhead. A multi-cast algorithm generates more control packets becauseit uses more routes (a multicast tree is formed frommultiple routes). Most multicast routing protocols forMANETs provide the shortest path (e.g., MAODV andRMAODV). During route discovery, it is faster to ob-tain the routing path for transmitting data, but this is

123


unreliable because of node mobility. The higher themobility is, the easier the routing paths will break. Inaddition, the power consumption of each node will fur-ther affect the data delivery ratio further. These factorswill decrease the data delivery ratio and require fre-quent routing maintenance.

In PMRP, we use GPS device to get the locationinformation of each node in advance and utilize thisinformation to improve routing efficiency. The multi-cast tree uses the routing path with the longest durationof time. The system lifetime of the multicast tree islonger than other multicast tree-based protocols. It notonly decreases the number of routing paths that breakbut also increases the packet delivery ratio.

6 Conclusions

In this paper, we proposed a power-aware multicastrouting protocol (PMRP) with mobility prediction forMANETs. In this scheme, we use the power-awaremetric to compute power consumption in advance. Wealso use the result to discover the efficient nodes anddecrease the numbers of control packets that need tobe broadcast. Then we use the global positioning sys-tem (GPS) to get the location information of the mobilenodes and utilize this information to calculate the LETbetween two connected mobile nodes. The destinationnode selects the routing path with the smallest LETand uses this smallest link expiration time as the RET.Each destination node collects several routing paths andselects the primary routing path with the longest RETfor data transmission. We also proposed a route main-tenance process to increase the packet delivery ratioand decrease the control overhead. Finally, the simula-tion results showed that PMRP outperformed MAODV[24] and RMAODV [2], especially in a high mobilityenvironment.

Acknowledgments This work was supported by the NationalScience Council of Republic of China under grants NSC-94-2213-E-324-025 and NSC-95-2221-E-239-052.

References

1. Bakhru, K. (2005). A seamless tracking solution for in-door and outdoor position location. In Proceedings of the16th IEEE international symposium on personal, indoor andmobile radio communications, Vol. 3, pp. 2029–2033, Sep-tember 2005.

2. Baolin, S., & Layuan, L. (2005). On the reliability ofMAODV in ad hoc networks. In Proceedings of the 2005IEEE international symposium on microwave, antenna,propagation and EMC technologies for wireless commu-nications, Vol. 2, pp. 1514–1517, August 2005.

3. Cheng, W.-H., Wen, C.-Y., & Feng, K.-T. (2006). Power-controlled hybrid multicast routing protocol for mobilead hoc networks. In Proceeding of the tenth IEEE 63thvehicular technology conference, pp.1087–1089, October2006.

4. Das, S. K., Manoj, B. S., & Murthy, C. S. R. (2002). Adynamic core based multicast routing protocol for ad hocwireless network. In Proceedings of the third ACM interna-tional symposium on mobile ad hoc networking and com-puting, pp. 24–35, June 2002.

5. Dommety, G., & Jain, R. (1996). Potential networkingapplications of global positioning system (GPS). TechnicalReport TR-24, Computer Science Department, The OhioState University, April 1996.

6. Djuknic, G. M., Richton, R. E. (2001). Geolocation andassisted GPS. Computer, 34(2), 123–125, February 2001.

7. Goel, A., & Sharma, A. K. (2006). A power managed basedmulticast routing protocol for mobile ad hoc network. InProceedings of the 2006 IFIP international conference onwireless and optical communications networks, pp. 5–9,April 2006.

8. Gopalsamy, T., Singhal, M., Panda, D., & Sadayappan, P.(2002). A reliable multicast algorithm for mobile ad hocnetworks. In Proceedings of the 2002 IEEE internationalconference on distributed computing systems, pp. 563–570,July 2–5.

9. Heinzelman, W. R., Chandrakasan, A., & Baladrishnan, H.(2000). Energy-efficient routing protocols for microsensornetworks. Proceedings of the 33rd Hawaii internationalconference on system sciences, Vol. 8, pp. 1–10, January2000.

10. Jaikaeo, C., Sridhara, V., & Shen, C.-C. (2006). Energy Con-serving Multicast for MANET with Swarm Intelligence. InProceedings of the 2006 international symposium on a worldof wireless, mobile and multimedia networks, pp. 7–11, June2006.

11. Kaplan, E. D. (1996). Understanding GPS: Principles andapplications. Boston, Artech Hourse: MA.

12. Kim, K. I., Mo, H. S., Baek, I. C., Shin, J. B., & Kim,S. H. (2003). Robust and cost-efficient group communica-tion using overlay multicast in mobile ad hoc networks. In2003 IEEE global telecommunications conference (GLO-BECOM), Vol. 2, pp. 713–717, December 2003.

13. Kin, D., Garcia-Luna-Aceves, J. J., & Obraczka, K. (2002).Power-aware routing based on the energy drain rate formobile ad hoc networks. In Proceedings of the IEEEinternational conference on computer communications andnetworks, pp. 565–569, October 2002.

14. Lee, S. J., Gerla, M., & Chiang, C. C. (1999). On demandmulticast routing protocol. In Proceedings of the 1999 IEEEWCNC, pp. 1298–1302, September 1999.

15. Liang, W. (2006). Approximate minimum-energy multicast-ing in wireless ad hoc networks. IEEE Transactions on Mo-bile Computing, 5(4), 377–387.

16. Liang, W., & Guo, X. (2006). Online multicasting for net-work capacity maximization in energy-constrained ad hoc

123


networks. IEEE Transactions on Mobile Computing, 5(9),1215–1227

17. Liao, W., & Jiang, M. Y. (2003). Family ACK tree(FAT): Supporting reliable multicast in mobile ad hoc net-works. IEEE Transactions on Vehicular Technology, 52,1675–1685, November 2003.

18. Liu, B. H., Tsai, M. J., & Ko, W. C. (2005). Dynamicalconstruction of a core-based group-shared multicast tree inmobile ad hoc networks. In Proceedings of the 19th interna-tional conference on advanced information networking andapplications, Vol. 1, pp. 90–95, March 2005.

19. Lynn, G. H. (2004). Robust multicasting using an under-lying link state unicast protocol. In Proceedings of the37th hawaii international conference on system sciences,pp. 7–11, January 2004.

20. Macker, J. P., & Corson, M. S. (1998). Mobile ad hocnetworking and the IETF. ACM SIGMOBILE Mobile Com-puting and Communications Reviews, 2(2), 9–14.

21. Pathan, A. S. K., Alam, M. M., Monowar, M. M., &Rabbi, M. F. (2004). An efficient routing protocol for mo-bile ad hoc networks with neighbor awareness and multi-casting. Proceedings of the E-Tech 2004, pp. 97–100, July2004.

22. Perkins, C. E., & Royer, E. (1999). Ad-hoc on-demand dis-tance vector routing. In Proceedings of the second IEEEworkshop on mobile computing system and application,New Orleans, LA, USA, pp. 90–100, February 1999.

23. Rappaport, T. S. (1995). Wireless communications: Princi-ples and practice. Upper Saddle River, NJ: Prentice-Hall.

24. Royer, E. M., & Perkins, C. E. (1999). Multicast operationof the ad hoc on-demand distance vector routing protocol. InProceedings of the ACM MOBICOM, pp. 207–18, August1999.

25. Scott, K., & Banmbos, N. (1996). Routing and channelassignment for low power transmission in PCS. In Proceed-ings of the international fifth ICUPC conference on univer-sal personal communications, Vol. 2, pp. 368–369, October1996.

26. Singh, S., Woo, M., & Raghavendra, C. S. (1998). Power-aware with routing in mobile ad hoc networks. In Proceed-ings of the ACM/IEEE international conference on mobilecomputing and networking, pp. 181–190, October 1998.

27. Toh, C. K. (2001). Maximum battery life routing to supportubiquitous mobile computing in wireless ad hoc networks.IEEE Communications Magazine, 39(6), 138–147.

28. Toh, C. K., Cobb, H., & Scott, D. A. (2001). Perfor-mance evaluation of battery-life-aware routing schemes forwireless ad hoc networks. In Proceedings of the 2001 IEEEinternational conference on communications, pp. 2824–2829, June 2001.

29. Vaishampayan, R., & Garcia-Luna-Aceves, J. J. (2004).Robust tree-based multicasting in ad hoc networks. In Pro-ceedings of the 2004 IEEE International conference on per-formance, computing, and communications, pp. 647–652,2004.

30. Xiong, X., Nguyen, U. T., & Nguyen, H. L. (2006). Pre-emptive multicast routing in mobile ad-hoc networks. InProceedings of the IEEE 2006 international conference onnetworking, international conference on systems and inter-national conference on mobile communications and learn-ing technologies. pp. 23–29, April 2006.

31. Yang, W.-L. (2005). Constructing energy-efficient multicasttrees with delay constraints in ad hoc networks. In Proceed-ings of the 19th international conference on advanced infor-mation networking and applications, pp. 414–419, March2005.

Nen-Chung Wang re-ceived the B.S. degree inInformation and ComputerEngineering from ChungYuan Christian University,Taiwan, in June 1990, andthe M.S. and Ph.D. de-grees in Computer Scienceand Information Engineer-ing from National ChengKung University, Taiwan, inJune 1998 and June 2002,respectively. From 2002 to2006, he was an AssistantProfessor in the Department

of Computer Science and Information Engineering, ChaoyangUniversity of Technology, Taiwan. Since August 2006, he joinedthe faculty of the Department of Computer Science and Infor-mation Engineering, National United University, Taiwan. He isa member of the Phi Tau Phi Society. His current research inter-ests include mobile computing, wireless networks, and paralleland distributed computing. Dr. Wang is a member of the IEEEComputer Society.

Jong-Shin Chen was bornin 1972. He received theB.Sc. and Ph.D. degreesin computer science fromFeng Chia University, Tai-wan, in 1996 and 2003,respectively. Currently, heis an assistant professor inthe Department of Gradu-ate Institute of Networkingand Communication Engi-neering, ChaoYang Univer-sity of Technology, Taiwan.

His research interests include mobile computing, capacity plan-ning, mobile agent, and wireless systems.

123


Yung-Fa Huang receivedthe College Diploma inElectrical Engineering fromNational Taipei University ofTechnology in 1982, M.S.degree in Electrical Engi-neering from National Hsin-hua University in 1987 andPh.D. degree in ElectricalEngineering from NationalChung Cheng University in2002. During 1987-2001,he was an instructor inChung Chou Institute ofTechnology. He is currently

an Associate Professor in Graduate Institute of Networking andCommunication Engineering, Chaoyang University of Technol-ogy. His current research interests include multiuser detectionin CDMA cellular mobile communication systems and wirelessnetworks.

Yu-Li Su received the B.S.degree in Electronics Engi-neering from Southern Tai-wan University of Technol-ogy, Taiwan, in June 2003,and the M.S. degree in Com-puter Science and Informa-tion Engineering from Cha-oyang University of Tech-nology, Taiwan, in June2005. His research interestsinclude mobile computingand wireless networks.

123

A Power-Aware Multicast Routing Protocol for Mobile Ad Hoc Networks With Mobility Prediction

Documents

Transcript of A Power-Aware Multicast Routing Protocol for Mobile Ad Hoc Networks With Mobility Prediction