Vous êtes sur la page 1sur 18

Computer Networks xxx (2011) xxxxxx

Contents lists available at ScienceDirect

Computer Networks
journal homepage: www.elsevier.com/locate/comnet

A cooperative multicast routing protocol for mobile ad hoc networks


I-Ta Lee a, Guann-Long Chiou a, Shun-Ren Yang b,
a b

Department of Computer Science, National Tsing Hua University, Hsinchu 300, Taiwan Department of Computer Science and Institute of Communications Engineering, National Tsing Hua University, Hsinchu 300, Taiwan

a r t i c l e

i n f o

a b s t r a c t
Many multicast routing protocols and algorithms have been proposed to support different group-oriented applications in mobile ad hoc networks. However, the conventional treebased and mesh-based multicast routing protocols generally suffer from frequent link breakage and high communication overhead, respectively. In this paper, we propose a cooperative multicast routing protocol, COMRoute, that incurs low communication overhead while guaranteeing reliable network connectivity. COMRoute utilizes cross-layer design by physical-layer cooperative transmission, MAC-layer clustering, and networklayer multicast routing. Specically, the physical-layer multi-node decode-and-forward cooperative transmission provides reliable transmission links. At MAC layer, nodes are classied into clusters. In each cluster, the cluster head, serving as a gateway, is responsible for inter-cluster transmission, while other nodes perform cooperative reception. Based on the clustered architecture, we design an on-demand source-based multicast routing protocol at network layer, which takes diversity into account for route establishment. Moreover, COMRoute implements a mechanism to mitigate the asymmetric cooperative transmission problem. Our simulation results show that COMRoute outperforms the representative multicast routing protocols in terms of trafc overhead, delivery ratio, and energy consumption. 2011 Elsevier B.V. All rights reserved.

Article history: Received 14 September 2010 Received in revised form 24 February 2011 Accepted 30 March 2011 Available online xxxx Responsible Editor: T. Melodia Keywords: Cluster Cooperative transmission Cross layer Hierarchical tree Multicast routing

1. Introduction In mobile ad hoc networks (MANETs), multicast routing has been widely applied to support different grouporiented applications (e.g., disaster recovery, video conferencing, and multimedia streaming) efciently. Owing to the plenty of advantages provided by multicast routing, many multicast routing protocols and algorithms have been proposed, including [18]. Specically, based on MANET multicast routing, in [9], instant messaging was employed to implement multi-user chat-rooms; in [10], mobile robots were implemented to detect hazards, to search and rescue, and to explore in hostile or inaccessible-by-human environments; in [11], a Doom-like
Corresponding author. Tel.: +886 3 5731212; fax: +886 3 5723694.
E-mail addresses: doug919@wmnet.cs.nthu.edu.tw (I-T. Lee), greencomet@wmnet.cs.nthu.edu.tw (G.-L. Chiou), sryang@cs.nthu.edu.tw (S.-R. Yang). 1389-1286/$ - see front matter 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.comnet.2011.03.019

augmented reality distributed game was implemented where the game scene can be a real, physical place, while tools like armors, medi-kits, etc. can be virtual objects within the real scene. According to the global topology of routing paths, these protocols can be classied into two categories: tree-based and mesh-based. Tree-based approaches [1,7,8] maintain a single loop-free route between the source and each receiver, while mesh-based approaches [3,5,4,6] construct multiple routing paths from a source to each destination and the nodes can deliver data copies through separate paths. Tree-based approaches have lowcontrol-overhead and bandwidth-efcient characteristics, but they suffer from the frequent link breakage caused by node mobility. Compared with tree-based approaches, mesh-based approaches are more resistant to the link breakage. However, they should pay for extra high maintenance cost and unnecessary bandwidth wastage. With the above respective defects of tree-based and mesh-based approaches, previous research results have shown that they

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

may not be suitable to be implemented in MANETs for realtime multimedia applications. To solve the problems caused by tree-based and meshbased approaches, it is required to design a new multicast routing protocol that incurs low communication overhead (as tree-based approaches) while guaranteeing reliable network connectivity (as mesh-based approaches). To achieve this goal, we incorporate the physical-layer cooperative communication technique in this paper. With cooperative communications, neighboring nodes overhearing a senders transmission may become cooperative partners and help the sender by forwarding the signals to the receiver. Through signal combination, this signicantly improves the link transmission reliability [1215]. Cooperative communications concept has been adopted in some proposed MANET routing protocols [1618], but all of these protocols are concentrated on unicast routing. With the assist of cooperative communication, we argue that a low-overhead tree-based multicast routing protocol can also provide a robust network connectivity comparable to mesh-based multicast routing protocols. In this paper, we propose a cooperative multicast routing protocol, COMRoute, which utilizes cross-layer design by integrating the physical-layer cooperative transmissions, Medium Access Control (MAC)-layer clustering and network-layer multicast routing. The physical-layer multi-node decode-and-forward cooperative transmissions provide reliable transmission links. At MAC layer, nodes are classied into clusters. In each cluster, the cluster head, serving as a gateway, is responsible for inter-cluster transmission, while other nodes perform cooperative reception. Based on the clustered architecture, we design an ondemand, source-based multicast routing protocol at network layer, which establishes a source-based multicast tree between clusters. Our protocol is the rst one to take diversity into account for route establishment. Specically, we select multicast routes that contain more cooperative partners to help cooperative transmissions. In this case, even under a highly dynamic environment, COMRoute can still construct reliable cooperative routing paths to achieve robust network connectivity. Moreover, our protocol also implements a mechanism to mitigate the asymmetric cooperative transmission problem, the problem that during the multicast route establishment, packets successfully transmitted from the source to destinations may be lost in the reverse direction. The main contributions of the paper are as follows: (1) This paper designs a complete cross-layer framework (from physical layer to network layer) for MANET multicast routing based on cooperative communications, aiming at striking a balance between control overhead and network connection robustness. To the best of our knowledge, this work is the rst to investigate the performance enhancement of physical-layer cooperative communications for MANET multicast routing. (2) We address the asymmetric cooperative transmission problem, which may occur frequently in ad hoc environments but has never been discussed in the literature.

(3) Our simulation results show that COMRoute efciently solves the frequent link breakage problem by using cooperative links, resulting in not only high network connection reliability, but also low communication overhead and energy consumption than traditional multicast routing protocols. We note that another multicast routing protocol Maximum-Residual Multicast Protocol (MRMP) [19] has recently been proposed in the literature, which obtains excellent performance in terms of similar output measures, i.e., trafc overhead, delivery ratio, and energy consumption. However, the objective of MRMP is to prolong the rst node failure time when network topologies and data trafc may change frequently in an unpredictable way, while the objective of our COMRoute is to guarantee reliable network connectivity with low communication overhead through cooperative communications. The rest of the paper is organized as follows. In Section 2, we describe the basic concept of multicast routing and cooperative communication. Section 3 introduces the considered physical-layer and MAC-layer enhancements of COMRoute. The multicast routing procedure of COMRoute is presented in Section 4. Section 5 examines the performance of COMRoute and compares it with a representative mesh-based multicast routing protocol, On-Demand Multicast Routing Protocol (ODMRP), and a simple tree-based multicast routing protocol. Finally, the paper is concluded in Section 6. 2. Basic concepts of ODMRP and cooperative communication In this section, we introduce the essential knowledge required in our work: ODMRP and cooperative communication. The former, having the best performance over lots of famous multicast protocols in most scenarios [20,21], is used as the performance benchmark for our method; the latter, a technique that improves reliability of wireless transmissions, is incorporated into our COMRoute, giving performance improvement, e.g., higher SNR and less energy consumption. 2.1. ODMRP: on-demand multicast routing protocol ODMRP, a mesh-based multicast routing protocol, is developed against the vulnerable multicast links in MANETs. Traditionally, multicast routing protocols are built on the tree-based topology, offering an efcient way to deliver data to multiple hosts. However, these tree-based protocols cannot resist link breakage problem caused by nodes moving around, resulting in the increase of multicast control overhead and the decrease of delivery ratio. Therefore, mesh-based ODMRP is invented. ODMRP contains four phases: (1) Join Request Phase. To start a multicast transmission, a source running ODMRP broadcasts a JOIN REQUEST message. Any node receiving this message will rebroadcast the message until all multicast receivers hear the request. Since nodes in the network may have more than one neighbors, they may receive

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

duplicate copies of JOIN REQUEST, each of which may have a different hop count. (2) Join Reply Phase. When a multicast receiver receives the rst JOIN REQUEST, it activates a timer. Before the timer expires, the receiver will continue to collect the same JOIN REQUEST from different paths. After the timer expires, the receiver selects the JOIN REQUEST with the minimum hop count for replying. If there are two or more JOIN REQUESTs having the minimum hop count to the source, the receiver will choose the one which arrived earliest to reply. The reply message is then transmitted along the reverse path of the request message, assigning the nodes on this path as forwarding group nodes. (3) Data Forwarding Phase. The sources send multicast streaming data along the forwarding group nodes to the destinations. (4) Maintenance Phase. Because nodes in MANETs may join/leave/move away from the multicast group, nodes in ODMRP need to maintain these paths through periodic JOIN REQUEST/REPLY cycles. That is, source nodes resend JOIN REQUEST messages after a certain period, and hosts need to reply the messages whenever they receive the requests if they want to stay in this multicast group. Therefore, ODMRP has the following features: (1) Forwarding Group. A forwarding group is a set of intermediate nodes on the shortest paths between each source/ destination pair, responsible for data forwarding and building the forwarding mesh. The forwarding mesh is a topology under which the nodes will have a link to each other when they are in each others transmission range. (2) Spare Routes. Because the forwarding mesh can provide spare routes for transmitting, ODMRP can resist link breakage problem when some routes are failed. (3) Ondemand Creation The routes are created only when the source wants to start transmission. (4) Soft State. No explicit leave messages are needed in ODMRP. When sources/ hosts want to leave the multicast group, they just stop transmitting JOIN REQUEST/JOIN REPLY messages. These features help ODMRP having better reliability and reducing control overhead. 2.2. Cooperative communication Cooperative communications are rstly developed against channel fading effect at physical layer in wireless ad hoc networks. Cooperative communications, exploiting the wireless broadcast nature and utilizing signal combination scheme, allow single-antenna mobile nodes form Multi-Input Multi-Output (MIMO) systems through sharing their antennas [12]. This is achieved by using a set of relay nodes, named helpers, which help a sender forwarding its packets. When the sender sends its packets, the helpers overhearing the transmissions will relay the overheard signals to the destination simultaneously. The receiver getting these signals then uses signal combination scheme to decode these superposition signals. Utilizing cooperative communications, nodes enjoy the following benets: (1) Signal Intensication. Through signal combining technique, the received signal can be intensied by receiving multiple copies of it. (2) Energy Conservation. Because the signal can be intensied by receiving multiple copies, each node using a smaller

transmitting power can still deliver data to its destination. It has been shown in [14,15] how cooperative transmission saves transmission power. (3) Spatial Diversity Gains. Since the received signals at a destination are acquired from different nodes in different positions, if some transmitting paths face highly intensive noise, signals from other paths still can be received by the destination. (4) Reliability. Spatial diversity offers extra reliability when the channel condition is poor. 3. COMRoute: physical layer and MAC layer enhancements This section presents the physical-layer and MAC-layer enhancements to support our proposed cooperative multicast routing protocol, COMRoute. We consider a mobile ad hoc network with n nodes, each of which equips a single antenna. Under this scenario, COMRoute is an on-demand source-based multicast routing protocol, in which sources establish routes only when they want to start a transmission. Fig. 1 shows the three-layer protocol architecture of COMRoute. In physical layer, we design a cooperative transmission scheme, CT, on top of the IEEE 802.11 physical layer. In MAC layer, a partner clustering algorithm, PCA, is introduced to partition the network into a hierarchical structure. Finally, in network layer, COMRoute is designed to support on-demand cooperative multicast routing. 3.1. Cooperative transmission scheme in physical layer COMRoute utilizes a multi-node decode-and-forward cooperative transmission scheme in physical layer based on a hierarchical architecture. In this architecture, nodes are separated into several clusters, each of which is constituted of one cluster head (CH) and many cluster members (CMs). The CH within a cluster, playing the role of a gateway, represents the cluster and conducts the multi-node decode-and-forward cooperative transmission between clusters. The other nodes within the cluster become the CMs to serve as the CHs virtual antennas, helping the CH receive and relay information cooperatively. These CMs may or may not hear each other and only need to directly communicate with the CH. Moreover, we assume that each node, depending on its current role in the network, can control their transmitting power between three power levels, PL1 ; P L2 , and Pcoop. PL1 is used to conduct intra-cluster transmissionor called layer-one (L1) transmission between CMs and their CHs, PL2 is used by CHs to

Fig. 1. Protocol architecture of COMRoute.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

communicate with each other in inter-cluster transmissionor called layer-two (L2) transmission, while Pcoop is for cooperative transmission between CMs and their CHs. Fig. 2 illustrates our considered transmission scheme. When CM1 of Cluster 1 has messages to send to Cluster 2, it transmits the messages to CH1 (the CH of Cluster 1) using PL1 for the L1 transmission, and the CH1 broadcasts the messages using PL2 for the L2 transmission. After that, the CMs in Cluster 2 receiving the messages cooperatively relay them to CH2 (the CH of Cluster 2) using Pcoop. Because we assume that CHs can combine signals through physicallayer cooperative transmission, the more CMs participating in the transmission, the less transmission power needed. Therefore, Pcoop can be expressed as

 10 log r

Pr ; Eo

where Eo is the received noise. Then, the received instantaneous SNR r is dened as an exponential random variable with rate 1. If the received instantaneous SNR r is less than  r a predened threshold, we say that the packet is failed to be received. Note that, for a cooperative transmission, the received instantaneous SNR at a CH (see CH2 in Fig. 2) can be calculated as

r coop r CH

NCM X i1

ri ;

Pcoop

a P L1
N CM

where rCH is the instantaneous SNR received from the neighboring CH (see CH1 in Fig. 2), and ri is the instantaneous SNR received from a CM i (see CM3 or 4 in Fig. 2). 3.2. Partner clustering algorithm in MAC layer In the two-layer architecture described in Section 3.1, it is clear that the CH election is the key to cluster stability and signicantly inuences network performance. Our partner clustering algorithm PCA modies the WCA in [24] to elect suitable CHs and partition the network into logically non-overlapped 1-hop clusters. Specically, our PCA adopts distance, relative speed and degree difference as metrics to periodically select proper CHs in a distributed manner, and nodes besides CHs will select one and only one CH to join, forming logically non-overlapped 1-hop clusters. Fig. 3 shows the three main steps of PCA. First, the nodes estimate their suitability for being a CH. After the estimation, a suitability dissemination mechanism is then exercised. Finally, the nodes with higher suitability scores than their neighbors are elected as the CHs. The details of these steps are elaborated on as follows. (1) CH suitability computation: In step 1 of Fig. 3, a node n in the network computes its suitability Sn for being a CH according to the following formula:

where NCM is the number of CMs, and a is a system parameter used to adjust a proper power level for the system. Eq. (1) is a simplied heuristic formulation. Since the objective of this paper is to investigate the relative performance differences of packet delivery ratio and control overhead, this simplied heuristic formulation is sufcient in the sense that it can clearly reect the power reduction due to cooperative communications. Note that other more sophisticated formulations can also be easily accommodated into our COMRoute design to replace Eq. (1) based on the interested performance metrics. To determine error packets, we rst need to compute the received instantaneous Signal-to-Noise Ratio (SNR). Following [22], our physical-layer model assumes that the wireless channel is a slow Rayleigh fading channel, under which the received power Pr is calculated through a simplied free space propagation model [23] as

Pr

P t k2 4p2 d
2

In (2), Pt is the transmission power, k is wavelength, and d is the distance between the transmitter and receiver. By using (2), we can derive the average received SNR  as r

Fig. 2. Multi-node decode-and-forward cooperative transmission between clusters.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

CH Suitability Score Computation

Start

(1)Computes Suitability Score

Suitability Score Dissemination

(2)BEACON Exchange

Cluster Construction and Maintenance

Has the Highest Suitability Score?

No

Yes

(3)Becomes a CH

No

Has a Cluster to Join?

(4)Waits Neighbors to Join

(5)Joins the Cluster

End

Fig. 3. Partner clustering algorithm PCA.

Sn

1 D n V n DN n

Vn

PK n q V nx V ni x 2 V ny V ni y 2 i1 Kn

where Dn is the average distance, Vn is the average relative speed, and DNn is the degree difference all between node n and its neighbors when node n uses PL1 . Dn, Vn, and DNn are further derived as follows. A node with a smaller average distance is likely to have nearer neighbors than other nodes. Consequently, the cluster constructed by this node is stabler, and this node is more suitable to be a CH. Dn can be formulated as

Dn

PK n q X n X ni 2 Y n Y ni 2 i1 Kn

where Kn is the number of 1-hop neighbors of node n (that is the number of nodes within ns direct communication range), (Xn, Yn) is the position of node n, and (X ni ; Y ni ) is the coordinate of the ith neighbor of node n. Similarly, the smaller the average relative speed Vn owned by a node n, the stabler the formed cluster. Vn is given as

where (Vnx, Vny) is the velocity of node n, and (V ni x ; V ni y ) is the velocity of the ith neighbor of node n. In the above Dn and Vn computations, the needed neighbor coordinate and speed information can simply be piggybacked in the MAC-layer BEACONs. When node n has received this information from all its neighbors, it can calculate Dn and Vn accordingly. The degree difference DNn represents the difference between the preferred cluster size r and the current number of node ns neighbors, where the preferred cluster size r is a system parameter. Clearly, a node with smaller DNn is more preferred by the system operator to become a CH. Moreover, by using r, we can control most clusters size in the system. DNn is expressed as

DNn jK n rj:
From the above discussion, we know that all the Dn, Vn, and DNn are preferred to be small. In this case, a

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

node with the largest Sn over its neighbors is suggested to be selected as a CH. (2) Suitability dissemination: In step 2 of Fig. 3, because we elect the CHs in a distributed manner, we need to propagate the suitability information across the network. At the beginning of each routing cycle, each node n updates its CH suitability score Sn and encapsulates it into a BEACON message. When the BEACON is broadcast, each 1-hop neighbor receiving this BEACON stores the Sn information. This stored Sn information is again broadcast with the next BEACON. The 1-hop and 2hop neighbor score information within each node can be maintained through this procedure. Based on this information, the CHs can thus be elected as follows. (3) Cluster construction and maintenance: In step 3 of Fig. 3, a node with the largest Sn within its 2-hop coverage area (i.e., among its 1-hop and 2-hop neighbors) becomes a CH. Note that using both 1-hop and 2-hop neighboring information to judge the suitability for being a CH is better than that using simply 1-hop neighboring information. This is because when a node judges its CH suitability through the 1-hop information, a large number of small clusters consisting of few or even a single node will be constructed. Consequently, the performance enhancement due to cooperative communication will be reduced. In step 4, the elected CH adds its CH information into its BEACON message and broadcasts it. The CHs neighbors receiving this message will join the cluster in step 5 and become CMs. The CMs also add their cluster information into the BEACON message, and thus the CH can maintain a cluster member list CML useful for the following routing process. Besides, PCA is executed periodically and a node without neighbors will form a cluster itself as a CH.

In suitability dissemination, if the BEACONs containing the suitability values got lost due to collision or some other issues, a node having a worse suitability value may announce itself as a CH. Consequently, two or more CHs in the same 2-hop coverage area appear and compete for CMs. However, nodes will consider that these CHs are for different clusters because they use the CH IDs to distinguish different clusters. That is, for each CH, it has its own cluster. Therefore, no clusters contain two or more CHs. 4. COMRoute: routing process This section describes the network-layer routing process of COMRoute. We rst dene the control packets and data structures which are required by the routing process. Then, we present the on-demand routing process in terms of three phases: Route Request (RREQ) phase, Route Reply (RREP) phase, and data transmission phase. Finally, we design a mechanism to overcome the asymmetric cooperative transmission problem. In this problem, packets successfully transmitted in one direction may be lost in the reverse direction. 4.1. Control packets In COMRoute, when a source has data to send to multiple destinations, it rst uses four kinds of control packets to create multicast routes: Route Request (RREQ) and Route Reply (RREP) are for intra-cluster communication, while Cluster Head Route Request (CHRREQ) and Cluster Head Route Reply (CHRREP) are for inter-cluster communication. A typical packet exchanging procedure illustrated in Fig. 4 is explained as follows. The source CM uses the RREQ (1) packet to request its CH to start an inter-cluster route discovery, and the CH receiving this packet then

Cluster 1

Cluster 3

RA D = 2 RM D = 2
(3)CH RREQ

RA D = 2 RM D = 1
(6)CH RREP

(2)CH RREQ

(4)CH RREQ CH RREQ CH RREP CH RREP

CH RREQ (7)CH RREP (5)CH RREP

1 RREQ (8)RREP

Source Cluster

Cluster 2

Destination Cluster

RA D = 3/ 2 RM D = 1
Cluster Head Cluster Member Source Destination
Fig. 4. Control packet exchanging procedure.

RA D = 2 RM D = 1

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

starts a route discovery by broadcasting the CHRREQ (2) packet. The neighboring clusters CMs successfully decoding this CHRREQ (3) packet cooperatively transmit it to their CH, and then the receiving CH combines the signals receiving from the last-hop CH and its CMs. The cooperative transmission of the CHRREQ packet is repeated for each cluster to build up the routing information. Once the destinations CHs receive the CHRREQ packet, they reply the CHRREP (57) packet to the source CH, activating the chosen routing paths. Finally, the source CH noties its source CM to start the data transmission by sending the RREP (8) packet. We further discuss the detailed information elds in these control packets. Since the RREQ packet is transmitted inside clusters, it only carries a little routing information, which contains the following elds:  Packet type shows the type of the packet.  Sequence number is uniquely assigned by the multicast hosts for identifying duplicated packets.  Time to Live (TTL) is the maximum number of hops that the packet can reach.  Multicast group address shows the address that represents the multicast group.  Source address shows the address of the source.  Last hop address shows the address of the packet sender in the last hop transmission. When a CH receives the RREQ packet from its CMs, it starts to construct inter-cluster routes by using a CHRREQ packet. In addition to the information included in RREQ, some new elds are added in the CHRREQ packet:  Hop count shows the number of clusters visited by the packet.  Route Average Diversity (RAD) records the average diversity of the visited clusters. Here the diversity of an intra-cluster communication is dened as the number of members (including CH) participating in this intra-cluster cooperative transmission. At the beginning of each routing cycle, the CHRREQ packet is generated by the source cluster with the RAD eld initialized to 0. During the traversal of the CHRREQ packet over different clusters, this eld will be recalculated and updated by the corresponding CHs as

consisting of few or even a single node (a high RMD), respectively. Fig. 4 shows a simple example of updating RAD and RMD. The source clusters CH starts the routing process by broadcasting a CHRREQ packet with PL2 . Members of the neighboring Cluster 1 receiving that packet cooperatively relay it to the CH. Therefore, the CH of Cluster 1 can know the number of CMs participating in this transmission, and computes the RAD by using Eq. (3). In addition to RAD, the RMD value is computed by comparing the diversity of the current cluster with the RMD recorded in the received CHRREQ packet. If the current diversity is smaller than the recorded one, then the CH replaces the RMD value with the current diversity. These two elds are updated hop-by-hop until reaching the destination. When the destination CH receives the CHRREQ, it will reply a CHRREP packet to establish the multicast route. Besides the elds in the RREQ and CHRREQ packets, the CHRREP packet introduces one more eld to indicate the selected route:  Forwarding Cluster Address indicates the next hop to forward the CHRREP packet. The next hop selection rule will be discussed in Section 4.3. As the CHRREP packet reaches the source CH, the source CH will transform the packet into the RREP packet, which has the same format as the RREQ packet. 4.2. Multicast routing table Each node running COMRoute has to maintain a multicast routing table (MRT) that aids the data forwarding. The structure of MRT is shown in Fig. 5. When a node receives a CHRREQ, it adds a record in MRT through backward learning. To be more specic, backward learning means to store the packet source in the Destination eld and the last hop in the next hop eld. The corresponding RAD, RMD, sequence number and time stamp are also collected by MRT. The Activate eld initialized as zero for each record and activated by a CHRREP packet indicates whether the next cluster is a forwarding cluster or not. The subsequent data packets will then follow the activated route recorded in MRT. The CHRREP Forwarder eld saves a designated CM which has the strongest SNR among the CMs that help the CHRREQ cooperative transmission. This CM will be useful for the CHRREP packet forwarding, which will be discussed later. 4.3. Cooperative multicast route creation and data transmission This subsection elaborates on the details of the route creation and data transmission procedures of COMRoute. The routing procedure of COMRoute contains three main phases: the RREQ phase, the RREP phase, and the data transmission phase. When a multicast source has data to send, it invokes the RREQ phase to discover the routes from the source cluster to the destination clusters. In the RREP phase, these inter-cluster routes are setup by assigning certain clusters as forwarding clusters. The CHs in the

RAD

RADlast Nhop 1 DIV current ; Nhop

where RADlast is the RAD recorded in the received CHRREQ packet, Nhop is the hop count until the current cluster, and DIVcurrent is the current diversity.  Route Minimum Diversity (RMD), similar to RAD, records the minimum diversity among the visited clusters. When CHs receive the CHRREQ packets, they will also update this information. Initially, this value is set to innity. Note that RAD and RMD are used during the RREP Phase, helping the destination CH select a route with more cooperative partners (a high RAD) and without clusters

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

Fig. 5. Multicast routing table.

selected forwarding clusters are responsible for data transmission using the links constructed through the physical-layer cooperative communication. The following subsections discuss these three phases in detail. 4.3.1. RREQ phase The RREQ phase invokes a route discovery procedure that consists of the following steps: (1) Intra-cluster RREQ transmission: If the multicast source is a CM, it asks its CH to start the inter-cluster route discovery by transmitting an RREQ packet. Otherwise, the procedure simply goes to step 2. (2) Inter-cluster CHRREQ broadcasting: When the source CH receives the RREQ packet from the source CM or the CH itself is the source, it broadcasts a CHRREQ packet using P L2 . (3) Cooperative transmission between partner clusters: Neighboring clusters receive the CHRREQ packet cooperatively, i.e., the CMs in neighboring clusters which successfully decode the packet will cooperatively transmit the packet to their CHs. These signals are combined by the CHs utilizing the physical-layer cooperative transmission. (4) Routing information updating: A neighboring CH which successfully decodes the CHRREQ packet judges whether the received CHRREQ packet is a duplicated one or not. When the CH receives a duplicated CHRREQ packet, it only updates the MRT. Otherwise, it searches the CML to check whether itself is a destination CH. If not, the receiving CH updates the MRT, recomputes RAD, RMD and other required elds for the CHRREQ packet and rebroadcasts this packet. (5) Phase transition: Steps 24 are performed repeatedly until the TTL of the CHRREQ packets is reached. Once a destination cluster is reached, the corresponding routing process moves into the RREP phase. Fig. 6 is an RREQ-phase example. In Fig. 6(a), the multicast source is in cluster C0, and two multicast group members specied as destinations are in clusters C7 and C8. A sequence of RREQ-phase operations are described as follows: (1) The multicast source CM in C0 transmits an RREQ packet to its CH to invoke the inter-cluster route discovery procedure. (2) The receiving CH broadcasts a CHRREQ packet to establish the inter-cluster routes. (3) Neighboring clusters C1, C2, and C3 successfully decode the packet cooperatively. (4) Since the packet is not a duplicate, the CHs check their CML and ensure that their clusters are not destinations. They update the routing information,

and then rebroadcast the packet. (5) Steps 24 are repeated until a destination cluster is reached. Fig. 6(b) illustrates the detailed CHRREQ paths from the source cluster to the two destination clusters. 4.3.2. RREP phase When a destination CH receives a CHRREQ packet, it waits a small amount of time to collect a few more CHRREQ packets and then transfers to the RREP phase. In this phase, the destination CH replies a CHRREP packet which assigns the forwarding clusters and establishes the routing paths. This forwarding cluster selection procedure performs as follows: (1) Minimum hop count information retrieving: The destination CH retrieves the minimum hop count information from all the received CHRREQ packets, and eliminates the routes whose hop counts are larger than the minimum hop count. (2) Diversity factor computation: Diversity factor listed below is a score that indicates the suitability of being a forwarding cluster, and it is used by the destination CH to select a forwarding cluster among the remaining routes

PNhop RMD i1 DIV i DF RMD RAD : N hop

There may exist other more sophisticated formulations that can be applied in COMRoute to express the diversity factor (DF). During the RREP phase, we prefer to select a routing path with not only a high RAD but also a high RMD. With a high RAD, the destination can select a routing path with more cooperative partners, increasing reliability during data transmission. On the other hand, with a high RMD, the destination can eliminate the routing paths containing clusters with few or even a single node, which are not benecial for cooperative transmission. In this case, it is reasonable to express DF as a proportional function of both RAD and RMD, as reected by the current Eq. (4). (3) Forwarding cluster assignment: The cluster with the largest diversity factor is chosen as the forwarding cluster and its address is lled in the Forwarding Cluster Address eld of a newly created CHRREP packet. The destination CH then broadcasts the CHRREP packet. (4) Inter-cluster CHRREP cooperative transmission: The neighboring clusters cooperatively receive the CHRREP packet. The cluster indicated in the Forwarding Cluster Address eld becomes the

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

C1
(4) (3)

C4

C7

C0

(2)

C2
(4)

C5

C8

(1)

(2)

(2)

C3
(4) (3)

C6

C9

Cluster Head Cluster Member Source Destination

C1

C4

C7

CH RREQ

CH RREQ

C0

CH RREQ

C2

duplcat i e

C5

CH RREQ

C8

CH RREQ

CH RREQ

CH RREQ

CH RREQ

C3

C6

C9

CH RREQ

CH RREQ

CH RREQ

Cluster Head Cluster Member Source Destination

Fig. 6. RREQ phase.

forwarding cluster, and its CH sets the Activate eld in its MRT to 1 and updates the corresponding routing information. (5) Route establishment: The forwarding clusters CH repeats steps 14, but this time it replaces the role of the destination CH to assign the next forwarding

cluster. This procedure repeatedly performs until the CHRREP packet reaches the source cluster. The corresponding routing path is thus constructed. Note that when determining the forwarding cluster, our proposed scheme does not only consider the information

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

10

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

given by the previous cluster. Instead, we use RAD and RMD, which collect the information over the routing path, to estimate the suitability of the path. In particular, RAD records the average diversity of the visited clusters traversed by the CHRREQ packet while RMD records the minimum

diversity among the visited clusters. Therefore, our algorithm gives a global solution. To explain the RREP phase more clearly, we take the previous RREQ phases destination C8 as an example. Fig. 7(a) shows the detailed operations: (1) There are four

C1 DIV=2

C4 DIV=3

C7 DIV=3

C0

C2 DIV=1

C5 DIV=4

14/3

C8

7/3 DIV=2

C3 DIV=2

C6 DIV=3

C9 DIV=3

Route 1

Cluster Head Cluster Member Source Destination

C1 DIV=2
CH RREP

C4 DIV=3
CH RREP

C7 DIV=3

C0

CH RREP

C2 DIV=1

C5 DIV=4

CH RREP

C8

DIV=2

C3 DIV=2

C6 DIV=3

C9 DIV=3

Cluster Head Cluster Member Source Destination

Fig. 7. RREP phase.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

11

candidate routes that can be selected by C8s CH: route 1 (C0 C3 C6 C9 C8), route 2 (C0 C2 C5 C8), route 3 (C0 C2 C4 C8), and route 4 (C0 C1 C4 C8). Among these routes, route 1 uses four hops to reach the destination C8 and the other routes use only three hops. Thus, C8 eliminates route 1 whose hop count is larger than the minimum hop count (i.e., 3 in this example). (2) C8 computes the diversity factors for the remaining routes. The factor for C4 in this example is 14/3 and that for C5 is 7/3. (3) We select C4, which has the largest diversity factor, as the forwarding cluster. Therefore, the destination C8 prepares a CHRREP packet and forwards the packet to C4. (4) The CH of C4 cooperatively receives the CHRREP packet from C8, and updates the routing information in MRT. (5) C4 becomes the role of C8 in step 1 and re-selects the next-round forwarding cluster. Fig. 7(b) shows the established routes. 4.3.3. Data transmission phase Once the source CH receives the CHRREP packet from all the destinations or a CHRREP timeout event occurs, it
Source Cluster

sends an RREP packet to notify the source CM to start the data transmission. Following the activated route in MRT, the data packets are cooperatively transmitted to the destinations. When the packets reach a destination CH, the CH communicates with the destination CM and transmits the data to it. Note that, we conduct the RREQ and RREP phases at the start of each routing cycle, and thus the routing paths are updated periodically. Besides, we do not need any extra control messages to refresh the routes. This method effectively reduces the control overhead and redundant packet transmissions. Note that COMRoute is a loop-free protocol. In this protocol, we consider two factors to build up multicast paths: shortest path to the source and diversity factors. The shortest path will be constructed from the destination cluster to the source cluster. If there are multiple shortest paths for the destination cluster to the source cluster, the CH will choose the path with the largest diversity factor. That is, a cluster has only one predecessor as its forwarding cluster. In this case, our COMRoute will construct a multicast tree of clusters. With such a tree topology, COMRoute is
Destination Cluster

CH RREQ
C1 CH C 1 C2 CH C 2 Cluster Head Cluster Member

Source Cluster

Destination Cluster

CH RREP
C1 CM 1 Cluster Head Cluster Member CM 2 CH C 1 CH C 2 C2

Fig. 8. Asymmetric cooperative transmission problem.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

12

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

therefore loop-free. Moreover, similar to many other existing on-demand source-based multicast routing protocols, our COMRoute uses a periodical updating procedure which renews the routing information per routing cycle to tackle the mobility issue. Clearly, the shorter the routing cycle, the higher the position precision and also the control overhead. Thus, the setting of the routing cycle depends on the tradeoff between position precision and control overhead. However, as an on-demand source-based protocol, COMRoute needs to periodically update the routing paths only when nodes have data to send. That is, it will not generate any control packets when nodes have no data to transmit. Therefore, low trafc results in low control overhead. 4.4. CHRREP forwarding: a mechanism to overcome asymmetric cooperative transmission During the routing process, it is possible that a CHRREQ packet is successfully transmitted in one direction while its CHRREP packet is lost in the reverse direction. In this case, the routing path corresponding to the lost CHRREP packet cannot be constructed. This is referred to as the asymmetric cooperative transmission problem in this paper, which has never been discussed in the previous work, e.g., [25,26]. Fig. 8 shows an example, where there are two immediate clusters, C1 and C2, on the CHRREQ forwarding path. In Fig. 8(a), although the CHRREQ packet cannot reach the CH of C2 (CHC 2 ) from the CH of C1 (CHC 1 ) directly, it can do this through the cooperative transmission by the two CMs of C2. As the routing process transfers into the RREP phase in Fig. 8(b), CHC 1 is, however, not only too far from CHC 2 but also has no CMs in the shaded area that can help transmit the CHRREP packet cooperatively. Therefore, the CHRREP packet will be lost and the route discovery procedure will fail, causing that the corresponding destination can not receive the data trafc in the subsequent data transmission phase. In fact, some CMs of a CHRREP forwarding cluster (e.g., CM1 and CM2 of C2 in Fig. 8(b)) may have the opportunity to help their CH forward the CHRREP packet. By using this observation, we can mitigate the asymmetric cooperative transmission problem which is caused by two consecutive clusters having unbalanced distribution of cooperative partners on their links. Fig. 9 shows how the chosen CM helps forward the CHRREP packet in the example of Fig. 8. CHC 2 now needs to forward the CHRREP packet to

C1. Because CHC 2 cannot forward this packet to CHC 1 directly or cooperatively through C1s CM(s), CHC 2 starts the following procedure to help C1 receive this CHRREP packet: (1) CHC 2 reads the CHRREP Forwarder eld in its MRT, which is updated in the RREQ phase, designating the recorded CM (i.e., CM1 in Fig. 9) as the CHRREP forwarder. We claim that since CM1 has previously forwarded the CHRREQ packet from CHC 1 to CHC 2 in the RREQ phase, it can now help its CH reply the CHRREP packet. (2) CHC 2 forwards the CHRREP packet to CM1 using the L1 transmission, and noties CM1 to forward this packet to C1. (3) CM1 sends the CHRREP packet to C1 using the L2 transmission, and any receiving CMs of C1 cooperatively relay this packet to CHC 1 . Notice that, this mechanism is exercised only when the following two conditions are satised: First, in the RREQ phase, between two immediate CHs, X and Y, X cannot transmit the CHRREQ packet to Y directly but can achieve this through CM cooperative transmission. Second, X is selected as the forwarding cluster in the corresponding RREP phase. 5. Simulation results This section shows the performance achieved by COMRoute, compared with ODMRP and the simple tree-based method (SimpleTree). In our simulation, we implement ODMRP and COMRoute; then, we use COMRoute to simulate SimpleTree, achieved by letting every cluster only contain one member. 5.1. Simulation environment Our results are collected through an event-driven simulator, which models the environment in terms of: network model, mobility model, and trafc model. These models are described as follows: (1) Network model: In our simulation model, 50200 nodes are randomly deployed on a 1000 m 1000 m square area, utilizing CSMA/CA without the retransmission mechanism for wireless transmission. The maximum transmission range of these nodes depends on whether L1 or L2 transmission is chosen, that is, nodes can reach 100 m when using PL1 and 300 m when using P L2 . Moreover, the network nodes use two SNR thresholds, Thrctrl and

Source Cluster

Destination Cluster

CH RREP
C1 CM 1
Cluster Head Cluster Member CHRREP Forwarder
Fig. 9. Losing CHRREP problem: solution.

C2 CH C 2

CH C 1

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

13

Thrdata, listed in Table 1 to determine the error packets for control and data packets, where Thrctrl and Thrdata use different modulation and coding schemes. Each scenario runs for 300 simulation seconds and the results are averaged over multiple runs. (2) Mobility model: All network nodes apply the random waypoint model, generated by CANUMobiSim [27]. In this model, nodes are uniformly distributed to an area initially, and each node moves to a randomly chosen destination with a velocity ranged from 0 to Vmax. Once reaching the destination, a node stays with a random period ranged from 0 to Trest,max, and then selects a new destination for next travel. (3) Trafc model: Source nodes in the network use Constant Bit Rate (CBR) trafc type, generating ve data packets per second. Each packet is composed of the data payload and its header with size Ppayload and Pheader respectively. We consider a multicast scenario with 1 source and 5 receivers, randomly chosen among the nodes in the network. Using only one node as source can simplify the simulation environment and at the same time allow us to concentrate on the relative performance comparison of reliability and control overhead. Our simulation experiments can be easily extended to accommodate multiple sources. In addition, the control packets with size Pctrl are used when a routing cycle Trouting is reached. 5.2. Simulation parameters Our considered simulation parameters are summarized in Table 2. Moreover, in Section 3.1, we mentioned that cooperative transmission can help save packet transmission energy, achieved by letting nodes use less power for packet transmission and using signal combining technology for decoding received packets. Eq. (1) suggests how nodes adjust their power to achieve this objective. To determine a suitable value of a in Eq. (1), we conduct a simple experiment, in which a CH is xed and various numbers of CMs are placed on the location that is RL1 away from the CH, where RL1 is the transmission range under PL1 . These CMs use Eq. (1) to decide their cooperative transmission power to forward messages to the CH. Fig. 10 shows the resulting success rates of the message forwardings. In this gure, a = NCM denotes the case that all CMs use P L1 and no power saving is achieved. Here, using cooperative communication, we can select an a to conserve the transmission power while reaching a success rate closer to that under a = NCM. Fig. 10 indicates that a = 1/0.5 = 2 results in a desired success rate, and saves more power than a = 1/ 0.3. Therefore, we choose a = 2 for the remaining

Table 2 Simulation Parameters Notation Simulation time Simulation area Number of nodes Basic rate Data rate Transmission power (range) Preferred cluster size COMRoute 300 s 1000 m 1000 m 50200 1 Mbps (BPSK) 11 Mbps (16-QAM) P L1 34 mW 100 m P L2 305 mW 300 m 3 1/5 1 3 25 m/s 10 s 2s 1024 bytes 52 bytes 90 bytes ODMRP

305 mW 300 m

r
Number of sources/ receivers Thrctrl Thrdata Vmax Trest,max Trouting Ppayload Pheader Pctrl

simulation experiments. Notice that when the number of CMs NCM is one, Eq. (1) is not applicable and the only CM simply uses P L1 to relay the messages. On the other hand, the setting of the preferred cluster size r mainly affects packet delivery ratio and control overhead. The larger the r, the less the control overhead and the smaller the delivery ratio. Thus, how to choose a proper r depends on which performance metric the network operator focuses on. Our experimental results indicate that the change of r (from 3 to 10) only has little impact on the two system performance metrics under the considered network environment setting in this paper. We found that the network topology, i.e., network density and node distribution, dominates the performance variations in these experiments. Specically, the suitability value Sn is mainly inuenced by the two network topology parameters, the average relative speed Vn and average distance Dn, under the considered simulation setting. That is, rather than r, these two parameters determine the formation of clusters. Therefore, for the demonstration purpose, we only consider r = 3 in the simulation experiments. However, when the candidate CHs have similar Vn and Dn, r will have a signicant impact on the cluster formation and will thus inuence the performance of packet delivery ratio and control overhead. 5.3. Simulation results In this section, we discuss the results obtained from COMRoute, ODMRP, and SimpleTree in four aspects: (1) trafc overhead, (2) packet delivery ratio, (3) energy consumption, and (4) mobility. Note that our COMRoute requires all nodes in the system to constantly update their coordinate and speed with each other. However, this update procedure has been widely employed by many previously proposed position-based MANET routing protocols [28,29]. In these position-based protocols, the coordinate and speed update information may simply be piggybacked in the regularly exchanged MAC-layer BEACON messages. Our protocol can share this information with them.

Table 1 SNR threshold. Notation Thrctrl Thrdata Packet Type Control packet Data packet Modulation BPSK 16-QAM Data rate 1 Mbps 11 Mbps

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

14

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

Fig. 10. Simulation results for adaptive power control.

Therefore, no extra communication should be spent in our scheme for these disseminations. In addition, the needed coordinate and speed information carried by BEACONs is relatively small in size compared with the BEACONs themselves. For example, the BEACON used by 802.11 AP occupies at least 40 bytes and our protocol may only require extra 6 bytes (4 bytes for coordinate and 2 bytes for speed) or less (by some compression scheme). In this case, this small amount of overhead can reasonably be ignored.

5.3.1. Trafc overhead Fig. 11 shows the total number of data packets existing in the network when delivering the same amount of data during the simulation period. In this gure, because of the characteristic of mesh-based topology, ODMRP generates duplicated data packets on the independent routes, resulting in the highest number of data packets. Our hierarchical tree-based COMRoute transmits packets only between clusters and thus uses less number of data packets, even than SimpleTree. Notice that, this amount difference increases as the number of network nodes

increases. In this case, COMRoute utilizes the bandwidth more efciently. Fig. 12 shows the number of control bytes transmitted per data byte delivered, pointing out how efciently the control packets are utilized in data packet deliveries. This gure reveals that SimpleTree has the highest control overhead, resulting from its tree-based topology and lowest packet deliverability. On the other hand, we can observe that COMRoute outperforms ODMRP. This is because compared with the hierarchical tree-based topology of COMRoute, ODMRP needs more control packets to maintain its mesh topology. Consequently, COMRoute utilizes the control packets more efciently, and hence it has better scalability to support more network nodes. 5.3.2. Packet delivery ratio Fig. 13 shows how packet delivery ratio varies with the number of network nodes. In this gure, we can nd that SimpleTree performs the worst because cooperative communication cannot be implemented over its tree-based routing path, resulting in serious link breakages. On the other hand, although ODMRP has a higher delivery ratio

Fig. 11. Total number of data packets.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

15

Fig. 12. Number of control bytes transmitted per data byte delivered.

Fig. 13. Packet delivery ratio.

Fig. 14. Total transmission energy consumption per data byte delivered.

than COMRoute when the network size is small, the delivery ratio of ODMRP eventually decreases as the number of

nodes increases. This is because of ODMRPs heavy overhead caused by frequent packet collisions in the dense

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

16

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

Fig. 15. Packet delivery ratio.

scenario. On the contrary, since COMRoute uses control and data packets effectively, its delivery ratio increases as the number of nodes increases under the considered range of network density. In summary, with the performance enhancement of cooperative communication, COMRoute has a comparable or even better delivery ratio than ODMRP which trades control overhead for higher deliverability. 5.3.3. Energy consumption Fig. 14 demonstrates the relationship between the total energy consumption per data byte delivered and the number of network nodes. COMRoute uses a hierarchical tree-based topology to eliminate redundant packet transmissions and applies cooperative transmission in physical layer. Therefore, we can see from the gure that the energy consumption of COMRoute is less than those of ODMRP and SimpleTree, especially in high density scenarios. 5.3.4. Mobility Fig. 15 shows how the packet delivery ratio changes when the mobility speed increases. In Fig. 15, we rst observe that the packet delivery ratio of SimpleTree decreases as the mobility speed of network nodes increases, while the other two protocols are insignicantly affected by node mobility. This is because the fragile tree topology of SimpleTree cannot afford the high mobility environment, resulting in frequent link breakages. On the contrary, under the hierarchical tree and mesh topologies, COMRoute and ODMRP provide more robust routing paths and both achieve acceptable packet delivery ratios even when the node mobility speed is high. 6. Conclusion In this paper, we proposed a cooperative multicast routing protocol, COMRoute, which utilizes cross-layer design by physical-layer cooperative transmission, MAC-layer clustering, and network-layer multicast routing. COMRoute took the diversity into account for route establishment, and thus achieved robust network connectivity. Moreover, our protocol also implemented a mechanism to mitigate

the asymmetric cooperative transmission problem ignored in the previous work. Simulation results showed that COMRoute efciently solves the frequent link breakage problem by using cooperative links, resulting in not only higher network connection reliability, but also lower communication overhead and energy consumption than the representative multicast routing protocols. As a nal remark, we identify possible directions for the future work of this paper.  In MANETs where energy is a scarce resource, selsh nodes may exist and there is no guarantee that any given protocols would be followed by all nodes managed by different authorities. In COMRoute, the existence of a selsh node will lead to the establishment of a longer or worse-diversity route, and/or the problem that clusters over the established paths may have less diversity factors than expected. To resolve these problems, different game-theoretical approaches can be employed to design a truthful mechanism for COMRoute. For example, we can use the Shapley Value [30] to allocate the payoff for each cluster. We allow an intermediate cluster to join the route if this relay can save aggregate transmission power. We measure the transmission power reduction and use it as a gauge of the payoff. The more power saved by the participation of the new cluster, the more payoff the cluster earns. By using Shapley Value, we can encourage clusters to cooperate.  On the other hand, our COMRoute borrows the overhearing concept to improve the network connection reliability. With signal combining of cooperative transmissions, COMRoute reduces single-node transmitting power. Moreover, COMRoute can also establish stable routes, avoiding energy wastage for data/control packet retransmissions. However, we realize that extra data receptions due to overhearing would be required and therefore would be an important factor affecting energy consumption. In this case, mechanisms to reduce the energy consumption from unnecessary overhearing merit further investigation.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx

17

Acknowledgments The authors thank the anonymous reviewers for their valuable comments, which have signicantly enhanced the quality of this paper. This work was supported in part by the National Science Council of Taiwan under Contract 98-2221-E-007-058-MY2, Contract 98-2221-E-009-059MY2, and Contract 98-2219-E-007-011-, and in part by Chunghwa Telecom.

References
[1] E.M. Royer, C.E. Perkins, Multicast operation of the ad-hoc ondemand distance vector routing protocol, in: Proceedings of the 5th Annual ACM/IEEE International Conference on Mobile Computing and Networking, MobiCom 99, ACM, New York, NY, USA, 1999, pp. 207218. [2] J. Garcia-Luna-Aceves, E. Madruga, A multicast routing protocol for ad-hoc networks, in: Proceedings of the 19th Annual Joint Conference of the IEEE Computer and Communications Societies (INFOCOM 99), vol. 2, 1999, pp. 784792. [3] S.-J. Lee, M. Gerla, C.-C. Chiang, On-demand multicast routing protocol, in: IEEE Wireless Communications and Networking Conference, 1999 (WCNC 1999), vol. 3, 1999, pp. 12981302. [4] X.-M. Sun, W.-J. Liu, Z.-Q. Zhang, Y. Zhao, Codmrp: cluster-based on demand multicast routing protocol, in: International Conference on Wireless Communications, Networking and Mobile Computing, 2006 (WiCOM 2006), 2006, pp. 15. [5] S.Y. Oh, J.-S. Park, M. Gerla, E-odmrp: enhanced odmrp with motion adaptive refresh, Journal of Parallel and Distributed Computing 68 (8) (2008) 10441053. [6] R. Biradar, S. Manvi, M. Reddy, Link stability based multicast routing scheme in MANET, Computer Networks 54 (7) (2010) 1183 1196. [7] C. Wu, Y. Tay, Amris: a multicast protocol for ad hoc wireless networks, in: IEEE Military Communications Conference Proceedings, 1999 (MILCOM 1999), vol. 1, 1999, pp. 2529. [8] J. Xie, R.R. Talpade, A. Mcauley, M. Liu, Amroute: ad hoc multicast routing protocol, Mobile Networks and Applications 7 (2002) 429 439. [9] D. Greene, D. OMahony, Instant messaging presence management in mobile adhoc networks, in: Proceedings of the Second IEEE Annual Conference on Pervasive Computing and Communications Workshops, 2004, pp. 5559. [10] S. Das, Y. Hu, C. Lee, Y.-H. Lu, An efcient group communication protocol for mobile robots, in: Proceedings of the 2005 IEEE International Conference on Robotics and Automation, 2005 (ICRA 2005), 2005, pp. 8792. [11] N. Reijers, R. Cunningham, R. Meier, B. Hughes, G. Gaertner, V. Cahill, Using group communication to support mobile augmented reality applications, in: Proceedings of the Fifth IEEE International Symposium on Object-Oriented Real-Time Distributed Computing, 2002 (ISORC 2002), 2002, pp. 297306. [12] A. Nosratinia, T. Hunter, A. Hedayat, Cooperative communication in wireless networks, IEEE Communications Magazine 42 (10) (2004) 7480. [13] A. Scaglione, D. Goeckel, J. Laneman, Cooperative communications in mobile ad hoc networks, IEEE Signal Processing Magazine 23 (5) (2006) 1829. [14] S. Cui, A. Goldsmith, A. Bahai, Energy-efciency of mimo and cooperative mimo techniques in sensor networks, IEEE Journal on Selected Areas in Communications 22 (6) (2004) 10891098. [15] Z. Zhou, S. Zhou, S. Cui, J.-H. Cui, Energy-efcient cooperative communication in a clustered wireless sensor network, IEEE Transactions on Vehicular Technology 57 (6) (2008) 36183628. [16] J. Zhang, Q. Zhang, Cooperative routing in multi-source multidestination multi-hop wireless networks, in: Proceedings of the 27th IEEE International Conference on Computer Communications (INFOCOM 2008), 2008, pp. 23692377. [17] L. Wang, K. Liu, An throughput-optimized cooperative routing protocol in ad hoc network, in: Proceedings of the 3rd IEEE International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications, 2009, pp. 1255 1258.

[18] H. Lichte, S. Valentin, H. Karl, I. Aad, L. Loyola, J. Widmer, Design and evaluation of a routing-informed cooperative mac protocol for ad hoc networks, in: Proceedings of the 27th IEEE International Conference on Computer Communications (INFOCOM 2008), 2008, pp. 18581866. [19] P.-C. Hsiu, T.-W. Kuo, A maximum-residual multicast protocol for large-scale mobile ad hoc networks, IEEE Transactions on Mobile Computing 8 (11) (2009) 14411453. [20] S.-J. Lee, W. Su, J. Hsu, M. Gerla, R. Bagrodia, A performance comparison study of ad hoc wireless multicast protocols, in: Proceedings of the 19th Annual Joint Conference of the IEEE Conference on Computer and Communications Societies (INFOCOM 2000), vol. 2, 2000, pp. 565574. [21] T. Kunz, Multicasting in ad-hoc networks: comparing maodv and odmrp, in: Proceedings of the Workshop on Ad hoc Communications, 2001, pp. 1018610190. [22] Y. Lin, J.-H. Song, V. Wong, Cooperative protocols design for wireless ad-hoc networks with multi-hop routing, Mobile Networks and Applications 14 (2009) 143153. [23] H. Friis, A note on a simple transmission formula, Proceedings of the IRE 34 (5) (1946) 254256. [24] W. Choi, M. Woo, A distributed weighted clustering algorithm for mobile ad hoc networks, in: Proceedings of the International Conference on Internet and Web Applications and Services/ Advanced International Conference on Telecommunications, 2006 (AICT-ICIW 06), 2006, p. 73. [25] P. Zhou, W. Liu, K. Xu, An energy-efcient multihop cooperative transmission protocol design for sensor networks, in: Annual IEEE Conference on Local Computer Networks, 2007, pp. 220222. [26] I.-T. Lin, I. Sasase, Distributed ad hoc cooperative routing in clusterbased multihop networks, in: Proceedings of the 2009 IEEE 20th International Symposium on Personal, Indoor and Mobile Radio Communications, 2009, pp. 26432647. [27] CANU Research Gro., CANU mobility simulation environment, 2001. <http://canu.informatik.uni-stuttgart.de/mobisim/index.html>. [28] M. Mauve, A. Widmer, H. Hartenstein, A survey on position-based routing in mobile ad hoc networks, IEEE Network 15 (6) (2001) 3039. [29] S. Giordano, I. Stojmenovic, L. Blazevic, Position based routing algorithms for ad hoc networks: a taxonomy, in: Ad Hoc Wireless Networking, Kluwer, 2001, pp. 103136. [30] J. Cai, U. Pooch, Allocate fair payoff for cooperation in wireless ad hoc networks using shapley value, in: Proceedings of the 18th International of IEEE Conference on Parallel and Distributed Processing Symposium, 2004, p. 219.

I-Ta Lee received the B.S. degree in computer science and engineering from Yuan Ze University, Taoyuan, Taiwan, in 2008 and the M.Sc. degree in computer science from National Tsing Hua University, Hsinchu, Taiwan, in 2010. In 2008, he joined the Wireless and Mobile Networking Laboratory, Department of Computer Science and Institute of Communications Engineering, National Tsing Hua University. Since 2010, he has been with Moxa, Taipei, Taiwan. His current research interests include the design and analysis of mobile telecommunications networks and performance modeling.

Guann-Long Chiou received the B.S. degree in computer science from National Tsing Hua University, Hsinchu, Taiwan, in 2006. He is currently working toward the Ph.D. degree with the Department of Computer Science, National Tsing Hua University. In 2008, he joined the Wireless and Mobile Networking Laboratory, Department of Computer Science and Institute of Communications Engineering, National Tsing Hua University. His current research interests include the design and analysis of mobile telecommunications networks and performance modeling.

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

18

I.-T. Lee et al. / Computer Networks xxx (2011) xxxxxx telecommunications networks, computer telephony integration, mobile computing, and performance modeling.

Shun-Ren Yang received the B.S. and M.Sc. degrees in computer science and information engineering and the Ph.D. degree from National Chiao Tung University, Hsinchu, Taiwan, ROC, in 1998, 1999, and 2004, respectively. From April 1, 2004 to July 31, 2004, he was appointed as a Research Assistant in the Department of Information Engineering, the Chinese University of Hong Kong. Since August 2004, he has been with the Department of Computer Science and Institute of Communications Engineering, National Tsing Hua University, Taiwan, where he is now an Associate Professor. His current research interests include design and analysis of mobile

Please cite this article in press as: I-T. Lee et al., A cooperative multicast routing protocol for mobile ad hoc networks, Comput. Netw. (2011), doi:10.1016/j.comnet.2011.03.019

Vous aimerez peut-être aussi