86 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO.

1, JANUARY 2010

Efficient Multicast Algorithms for Multichannel

Wireless Mesh Networks
Guokai Zeng, Student Member, IEEE, Bo Wang, Student Member, IEEE,
Yong Ding, Student Member, IEEE, Li Xiao, Member, IEEE, and Matt W. Mutka, Senior Member, IEEE

Abstract—The wireless mesh network is an emerging technology that provides high quality service to end users as the “last mile” of
the Internet. Furthermore, multicast communication is a key technology for wireless mesh networks. Multicast provides efficient data
distribution among a group of nodes. However, unlike other wireless networks, such as sensor networks and MANETs, where multicast
algorithms are designed to be energy efficient and to achieve optimal route discovery among mobile nodes, wireless mesh networks
need to maximize throughput. This paper proposes two multicast algorithms: the Level Channel Assignment (LCA) algorithm and the
Multichannel Multicast (MCM) to improve the throughput for multichannel and multi-interface mesh networks. The algorithms build
efficient multicast trees by minimizing the number of relay nodes and total hop count distances of the trees. The algorithms use
dedicated channel assignment strategies to reduce the interference to improve the network capacity. We also demonstrate that using
partially overlapping channels can further diminish the interference. Furthermore, additional interfaces help to increase the bandwidth,
and multiple gateways can further shorten the total hop count distance. Simulations show that those algorithms greatly outperform the
single-channel multicast algorithm. We also observe that MCM achieves better throughput and shorter delay while LCA can be realized
in distributed manner.

Index Terms—Wireless mesh networks, multicast, multichannel, multi-interface, channel assignment.

1 INTRODUCTION

T HE Wireless Mesh Network (WMN) is an emerging

paradigm for the next-generation wireless Internet. In
such networks, most of the nodes are either stationary or
Internet. For example, a large number of users may watch the
FIFA World Cup on the Internet. The gateway that helps to
connect the mesh network with the Internet can effectively
minimally mobile and do not have power constraints. multicast the data packets to those users.
Compared with their single-hop counterpart, wireless LANs, Efficient multicast protocols in WMNs cannot be
WMNs are self-organized with the nodes automatically achieved by adopting or slightly modifying the multicast
establishing ad hoc networks and maintaining their con- protocols for other types of multihop wireless networks.
nectivity. This provides improved reliability as well as larger Unlike mobile ad hoc networks or wireless sensor networks,
coverage and reduces equipment cost. Being used for the last route recovery or energy efficiency is not the major concern
mile for extending or enhancing Internet connectivity, for mesh networks due to the limited mobility and the
commercial deployments of WMNs are already in the works, rechargeable characteristic of mesh nodes. Moreover,
such as MIT Roofnet [1] and Seattle Wireless [2]. supporting potential major applications, such as Video On
Mesh networks are characterized by the use of multiple Demand, poses a significant challenge for the limited
channels and multiple interfaces to improve system through- bandwidth of WMNs. Thus, it is necessary to design an
put. Recent research has focused on how unicast routing effective multicast algorithm for mesh networks.
assigns channels to different wireless interfaces to improve Traditional multicast protocols for wireless networks
system throughput in WMNs. However, multicast commu- assume that each node is equipped with one interface. A
nication, which intends to transmit the packets from the mesh network provides the nodes with multiple interfaces
source to a set of nodes, draws less attention in the literature that can be used to improve the throughput substantially.
of mesh networks. We believe that efficient multicast, which However, channel assignment is subject to the number of
cannot be readily achieved through combined unicast or available channels and interfaces, the network topology,
simplified broadcast, is essential to wireless mesh networks the communication requests, and other factors. Interference
and is worthy of thorough investigation. It is often necessary cannot be completely eliminated due to the limited number
for a portion of end users to retrieve data packets from the of available channels. An inappropriate channel assign-
ment strategy will result in throughput reduction due to
the multichannel hidden terminal problem [3], disconnec-
. The authors are with the Department of Computer Science and tion of the topology [4], or unfair bandwidth allocation to
Engineering, Michigan State University, 3115 Engineering Building,
East Lansing, MI 48824. various users [5].
E-mail: {zengguok, wangbo1, dingyong, lxiao, mutka}@cse.msu.edu. In this paper, we aim to design a multicast protocol for
Manuscript received 10 June 2008; revised 8 June 2009; accepted 2 Mar. 2009; mesh networks that has the following characteristics: 1) it
published online 10 Mar. 2009. improves the system throughput by allowing simultaneous
Recommended for acceptance by P. Mohapatra. close-by transmissions with multichannels and multi-inter-
For information on obtaining reprints of this article, please send e-mail to:
[email protected], and reference IEEECS Log Number TPDS-2008-06-0223. faces, and 2) it assigns all the available channels to the
Digital Object Identifier no. 10.1109/TPDS.2009.46. interfaces instead of just the nonoverlapping channels.
1045-9219/10/$26.00 ß 2010 IEEE Published by the IEEE Computer Society
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 87

We propose a Level Channel Assignment (LCA) algo- To simplify the system model, we consider the network
rithm and a Multichannel Multicast (MCM) algorithm to as a graph G ¼ ðV ; EÞ, where V represents the set of
improve throughput for multichannel and multi-interface gateways and mesh routers and E represents the physical
mesh networks. The algorithms first build a multicast links among neighboring nodes (the node refers to the mesh
structure by minimizing the number of relay nodes and hop router or the gateway in the subsequent sections). We
count distances between the source and destinations, and assume that each node has the same fixed communication
use dedicated channel assignment strategies to improve the range, that is, if node u can transmit directly to node v (and
network capacity by reducing the interference. vice versa), there is a link ðu; vÞ in E.
Our design builds a new multicast backbone—“tree The number of available channels is limited in the
mesh,” which partitions the mesh routers into different current network protocols. In addition, each node is able to
levels based on the Breadth First Search (BFS), and then be equipped with kðk  2Þ Network Interface Cards (NICs),
heuristically assigns channels to different interfaces. Tree- any of which can be tuned to any available channel.
based multicast is well established in wireless networks for Multichannel and multi-interface characteristics enable
its data forwarding efficiency over other types of ap- more concurrent transmissions. When one NIC is transmit-
proaches at the expense of low robustness. However, unlike ting or receiving packets on one channel, another NIC on
MANETs, WMNs are normally considered stationary and the same node is able to undertake transmission on another
always put throughput maximization as the first priority. different channel at the same time. The value of k usually
Thus, tree-based multicast is suitable for WMNs since the equals to 2, 3, or 4 due to economical reasons. In this paper,
topology change is not a major concern in WMNs. we first consider the situation that each node has two
We also illustrate that the use of partially overlapping interfaces, then we apply our algorithms to more interfaces.
channels can further improve the throughput. Simulations
show that our algorithms greatly outperform the single- 2.2 Measuring Partial Overlap
channel multicast algorithm. We observe that MCM achieves To improve the throughput of WMNs, many studies have
better throughput and shorter delay while LCA can be been conducted on how to assign orthogonal channels to
implemented in a distributed manner. adjacent wireless links to minimize interference. It is known
The rest of this paper is organized as follows: Section 2 that 802:11b=g and 802:11a provide 3 and 12 nonoverlap-
describes the system model and the design consideration. ping channels, respectively. Although 802:11a provides
Section 3 proposes an intuitive algorithm, the LCA algo- more nonoverlapping channels than 802:11b=g, it has
rithm, which is easy to implement but has drawbacks. several drawbacks. Because 802:11a works on a higher
Section 4 introduces the MCM algorithm to build a more frequency spectrum (5 GHz) than 802:11b=g (2 GHz), it is
efficient multicast structure; this is followed by a description more difficult to penetrate walls and other obstructions, and
of how to assign channels on it. Several companion thus 802:11a has a shorter range. In addition, the interfaces
mechanisms for our protocol are presented in Section 5. and access points for 802:11a are more costly. As a result,
Section 6 presents simulation results. Section 7 surveys the 802:11b=g is more commonly used.
related work, and the last section concludes this paper. Previous channel assignment algorithms for 802:11b=g
only use three nonoverlapping channels: 1, 6, and 11. In
these studies, a binary interference model is usually
2 SYSTEM MODEL assumed, that is, if two links are within interference range
We start from the underlying network model by introdu- of each other, they will interfere with each other if they are
cing some basic terminology and the partial channel conflict on the same channel, and otherwise not. However, the
phenomena, which is followed by design considerations for interference can be further reduced by using the partially
multicast algorithms in WMNs. overlapping channels too, that is, by using any channel
from 1 to 11 in the channel assignment.
2.1 Basics Through experiments, we observe that the interference
Mesh networks are composed of three types of nodes: between two links depends on both their physical distance
gateways (access points), mesh routers, and mesh clients. and channel separation [6]. Unlike the traditional inter-
Gateways enable the integration of WMNs with various ference model, the interference range is no longer a constant.
other networks, including the Internet. As dedicated devices Instead, it varies with the channel separation. Let Ic be the
provide stable high throughput for mesh clients, mesh interference range of two links with channel separation c.
routers have minimal mobility and form the mesh backbone. That means, when the channel separation of two links is c,
In order to further improve the flexibility and capacity of they will interfere with each other if their distance is less than
WMNs, the mesh routers are often equipped with multiple Ic , and otherwise not. For example, I0 ¼ 2R, which means the
wireless interfaces. As a result, two transmissions of two same channel can be used on two links without any
nearby pairs can be simultaneously scheduled if nonoverlap- interference only when they are over twice the transmission
ping channels are assigned. Mesh clients are usually end range away. In [6], [7], [8], experiments have been done to
users, such as laptops and PDAs, which access the Internet measure the interference between two wireless AP-Client
through the mesh routers so that the mesh clients are usually links with different distances and channel separations.
within one hop of the mesh routers. Since the multicast Definition 1. Interference Factor is defined as the ratio of the
packets are always relayed among the mesh backbone, we interference range to the transmission range. We use t to
only consider how to transmit the packets to multiple mesh represent the Interference Factor when the channel separation
routers; then packets will be forwarded one more hop to the of two links is t.
corresponding mesh clients that desire to receive the packets.
88 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO. 1, JANUARY 2010

Fig. 1. Experiment.

We perform experiments to measure the interference

factor between two wireless links. We use four laptops
with Netgear WAG511 PC Cards, each two of which
construct a separate wireless link as shown in Fig. 1. We
evaluate the interference between two links by comparing Fig. 2. Interference factor versus channel separation.
the total throughput when both links are active and the
sum of each link’s throughput when the other link is disconnection and exacerbation of multichannel hidden
turned down. The length of each link is fixed at 5 m. Linux terminal problems, which reduces the system throughput.
kernel with Madwifi is used to drive the network cards. Therefore, both efficient multicast structure and effective
The two end nodes of the link work on the same channel. channel assignment play important roles in mesh network
We configure the two links with different channels, and we multicast.
vary the distance d between the two links to find out the
interference range.
We found a similar trend in our experiments, that is,
3 LEVEL CHANNEL ASSIGNMENT ALGORITHM
interference range decreases with the increase of channel A common method for multicast is to build a multicast tree,
separation. The results of these experiments are shown in where the source node is usually the gateway. In this paper,
Fig. 2. Therefore, if we can fully utilize these partially we first propose the LCA algorithm, which can be achieved
overlapping channels, we can further decrease the total by the following steps.
interference in the network, and thus improve the net- First, the nodes obtain their level information. The BFS is
work throughput. used to traverse the whole network. All the nodes are
partitioned into different levels according to the hop count
2.3 Design Consideration
distances between the source and the nodes.
Routing protocols do exist to offer efficient multicasting
service for conventional multihop wireless networks, such as Definition 2. If node a (in level i) and b (in level i þ 1) are
MANETs and wireless sensor networks. Since the nodes within each other’s communication range, then a is called the
become increasingly mobile or the networks only have scarce parent of b, and b is called the child of a.
resources such as power constraints and limited computing
ability, most previous work pays much attention to energy Second, we build a multicast tree based on the node level
efficiency and how to build the multicast structure without information. Initially, the source and all the receivers are
knowing the global topology. As a result, the multicast included in the tree. Then, for each multireceiver v, if one of its
structure should be distributedly constructed, energy effi- parents is a tree node, then connect it with that parent, and
cient, and should take care of the topology change as well as stop. Otherwise, randomly choose one of its parents, say fv , as
group member management, which may conflict with relay node on the tree, and connect v and fv . Afterwards, we
maximizing the throughput of the network to some extent. try to find out the relay node for fv recursively. This process
However, since mesh networks are deployed to provide repeats until all the multireceivers are included in the
last-mile Internet access for enterprises or communities, the
multicast tree. Algorithm 1 gives the detail.
throughput and the network capacity are the major concerns.
Deployed at fixed locations, mesh routers have limited Algorithm 1. Multicast tree construction for LCA
mobility. Furthermore, they are computationally powerful Data: M: multi-receivers; s: source node;
and do not rely on battery power compared with their Result: T : multicast tree
counterparts in MANETs or sensor networks, which help to V ðT Þ ¼ M [ fsg; EðT Þ ¼ ;;
achieve sufficient network capacity to meet the requirement for 8 node v 2 M do
of applications such as audio or video sharing among end
p ¼ v;
users. Thus, we need to create a multicast structure that aims
to deliver the packets rapidly to the multireceivers (multi- while none of p’s parents is included in V ðT Þ do
receivers are defined as the multicast group members except Randomly select one of p’s parents, say fp .
for the source node) without worrying about the energy V ðT Þ ¼ V ðT Þ [ ffp g;
consumption and topology changes. EðT Þ ¼ EðT Þ [ fðp; fp Þg;
Moreover, equipping the mesh routers with more than p ¼ fp
one wireless interface could further improve the network end
capacity. The assignment of channels to interfaces on the EðT Þ ¼ EðT Þ [ fðp; fp0 Þg (fp0 is the parent of p, and it is a
multicast structure is also essential to throughput optimiza- tree node)
tion. Inappropriate channel allocation will lead to topology end
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 89

Fig. 3. An example for LCA and tree mesh. (a) Network topology, (b) multicast tree, (c) channel assignment, and (d) tree mesh.

Next, the tree nodes decide their channel assignment channel at the same level. For example, in Fig. 3c, since g is in
with the level information. the transmission range of both c and d, there will be
interference when c and d use the same channel. Second,
1. The source node (level 0Þ only uses one interface,
when the number of available channels is more than that of
which is assigned channel 0. This interface is
the levels, some channels will not be utilized, which is a
responsible for sending packets to the tree nodes in
waste of channel diversity. Third, the channel assignment
level 1.
2. The internal tree node in level iði  1Þ uses two does not take the overlap property of the two adjacent
interfaces: one is assigned channel i  1, which is channels into account. As we know, 8i, channel i and channel
used to receive packets from the upper level; the i þ 1 are adjacent in frequency, so they partially interfere
other is assigned channel i, which is used to forward with each other. Thus, the channel i for level i still has some
packets to tree nodes at level i þ 1. inference effect with the channel i þ 1 for level i þ 1.
3. The leaf in the level iði  1Þ uses two interfaces: one
uses channel i  1 to receive the packets from level 4 MULTICHANNEL MULTICAST ALGORITHM
i  1, the other uses channel i to forward the packets
to the mesh clients within its communication range To further improve the system throughput, we propose an
that desire to receive the packets. MCM algorithm to minimize the number of the relay nodes
One example is shown in Fig. 3, where node s is the source and the hop count distances between the source and the
and e; f; g are the multireceivers. Initially, {s, e, f, g} are destinations, and further reduce the interference by
included in the multicast tree. At first, since none of exploiting all the partially overlapping channels instead of
g’s parents are tree nodes, randomly select one parent d as a just the orthogonal channels.
tree node and connect g with d. We then choose d’s parent b as 4.1 Multicast Structure Construction
a tree node and connect d with b. Since b’s parent s is a tree
Following the design constraint of WMNs, we aim for a
node, we connect b with s and stop the process for including g
multicast protocol for WMNs, which includes two primary
in the multicast tree. Next, we start from the second
procedures. The first is to build an effective multicast
multireceiver e. Connect e with its parent b and stop, since b
structure, which is detailed in this section, and the second
is already a tree node. Similarly for the third multireceiver f,
tries to allocate channels for minimizing interference in the
we connect f with c; c with a, and then a with s. Now the tree
construction is complete since all the receivers are connected next section.
to the tree. The constructed multicast tree is shown in Fig. 3b.
4.1.1 Broadcast Structure
We can see that in the tree, level 0 ¼ fsg, level 1 ¼ fa; bg,
Some previous work treats broadcast and multicast in a
level 2 ¼ fc; d; eg, and level 3 ¼ ff; gg. Thus, we get the
channel assignment in Fig. 3c, where the number above the different way. Actually, when all the nodes are multi-
node represents the channel for receiving and the number receivers, the multicast problem becomes the broadcast
below the node represents the channel for sending. problem. We can say that broadcast is a special case of
The LCA algorithm has two advantages: simple imple- multicast. In order to focus on the basic idea of MCM, we first
mentation and throughput improvement. It only needs one consider the situation that all the nodes are the multi-
BFS of the network at the beginning, and it creates receivers. We then detail how to trim off those unnecessary
the multicast tree by connecting the multireceivers with branches based on the broadcast structure when the multi-
the nearest tree nodes. The tree nodes then can decide the receivers are only a portion of the nodes. The broadcast
channels by themselves according to the level information, structure in the mesh network is built by the following steps.
which can be realized distributedly. At the same time, the The first step is realized by BFS, which is similar with
use of multiple channels reduces the close-by interference the LCA algorithm. After the BFS traversal, all the nodes are
and allows more simultaneous transmissions. divided into different levels. We then delete the edges
However, there is still potential for the LCA algorithm to between any two nodes of the same level, with which we
improve system throughput. First, LCA cannot diminish the get the elementary communication structure—“tree mesh.”
interference among the same levels since it uses the same Figs. 3a and 3d give an example of the original network
90 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO. 1, JANUARY 2010

Fig. 4. Relay node search example.

topology and its corresponding tree mesh. We use BFS to Result: R: the set of the relay nodes in level i
build the tree mesh for the following reasons: R ¼ ;;
While Sj ! ¼ ; do
1. With the hop count distance increasing between the
In TG ðiÞ, compute the number of parents of each node
sender and the receiver, the intraflow contention
in Sj , and compute the number of children of each
exacerbates. Moreover, shorter hop count distance
node in Si ;
means shorter transmission delay. Minimizing the
delay is also important in WMNs, thus we build a Find vi1 ; vi2 ; . . . . in Sj with the minimal number of
shallow tree by BFS, which reduces the total hop parents;
count distances from the source to the receivers. Among the parents of vi1 ; vi2 ; . . . ., find tf with the
2. BFS guarantees that if two nodes are not at the same maximal number of children;
level or the adjacent levels, they are at least two hops R ¼ R [ ftf g;
away. Hence, when considering channel assignment, Si ¼ Si  ftf g;
the two nodes may use the same channel since they The children of tf record tf as their relay node;
are unlikely to interfere with each other. Sj ¼ Sj  fthe children of tf g;
3. The time complexity of BFS is OðjV j þ jEjÞ, whose end
cost is much less than other broadcast or multicast We can see that identifying the minimal number of relay
tree construction algorithms. nodes at level i is equivalent to selecting the minimal number
In the second step, we identify the minimal number of of nodes at upper node set of TG ðiÞ that can cover all the nodes
relay nodes that form the broadcast tree. Using more relay of lower node set. In fact, it is a variation of the set-cover
nodes means more transmissions in the network. Because the problem, which has been proved NP-complete. We devise an
number of available channels is limited by current technical approximation algorithm, which is detailed in Algorithm 2.
conditions, more transmissions would result in more inter-
1. Some parents are considered as relay candidates if
ference and incur more bandwidth cost. Thus, minimizing
one of their children has the minimal number of
the multicast tree size helps to improve the throughput. In the
parents.
tree mesh, one node could have more than one parent. The
2. Among the relay candidates, we choose one node
purpose of this step is to identify the only parent (we call it a
that has the maximal number of children. The reason
relay node here) for a node that has more than one parent so
is that given the fixed number of nodes at the level
that the number of relay nodes is minimal.
i þ 1, the more children a relay node can forward
A top-down approach, i.e., from level 0, level 1 to the
packets to, the less number of relay nodes we will
lowest level, is used to identify the relay nodes. Suppose we
need at level i.
have discovered the relay nodes in level 0, level 1; . . . ; level
3. We remove the relay node and its children, and
i  1; now we study how to find out the relay nodes in level
repeat the above process until all the nodes at level
i. We can see that fewer relay nodes will result in less traffic
i þ 1 are removed.
flows in the network, which means less local interference.
We use a simple example to further explain this algorithm.
Thus, our objective is to identify the minimal number of
Fig. 4a gives a (i; i þ 1) subtree mesh, from which we can
relay nodes in level i that can communicate with all the
compute the number of parents of each node in level i þ 1.
nodes in level i þ 1.
The nodes 1, 5, and 7 have the minimal number of parents
Definition 3. Given a tree mesh TG ; TG ðiÞ is a subgraph of TG (one parent), and their parents are nodes a; c, and d. The
and consists of only the nodes at level i and level i þ 1 of TG . numbers of children of nodes a; c, and d are three, two, and
This subgraph TG ðiÞ is called ði; i þ 1Þ subtree mesh. In two, respectively. Since node a has the maximal number of
addition, the set Si consisting of the nodes from level i is called children, a is chosen as a relay node.
the upper node set of TG ðiÞ, and the set Sj consisting of the We then remove a and its child nodes 1, 2, and 3 from the
nodes from level i þ 1 is called the lower node set. subtree mesh. In the new subtree mesh, which is shown in
Fig. 4b, the nodes 5 and 7 have the minimal number of parents,
and their parents are nodes c and d. We randomly choose c as
Algorithm 2. Relay node search in level i algorithm one relay node since c and d both have two children.
Data: TG ðiÞ: ði; i þ 1Þ subtree mesh; Si : nodes in level i; Sj : Afterward, we remove c and its children, then get the
nodes in level i þ 1 subtree mesh shown in Fig. 4c. Similarly, with the process
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 91

Fig. 5. A multicast structure example. (a) Network topology and (b) multicast tree.

above, we select node d as a relay node. After removal of The goal of the algorithm is to discover the minimal
node d and its children, the level i þ 1 is empty, thus the number of relay nodes needed to construct a multicast tree.
algorithm stops. Finally, nodes a; c, and d are chosen as The search process starts from the bottom to the top. We use
relay nodes at level i, which is shown in Fig. 4d. a boolean variable “c[v]” for any node v to represent that v
Algorithm 2 is superior to the Greedy Set Cover is either a multireceiver or a relay node if c[v] is true. At
algorithm [9] by introducing step 1. We observe that if a each step, we intend to minimize the number of relay nodes
node has just one parent, the parent has to be selected as a at the upper level, which can cover all the multireceivers
relay node, while that greedy algorithm recursively selects and relay nodes at the lower level. The process is similar
the node with the maximal number of children in the with the broadcast structure, except that we do not require
remained graph. For the above example, the greedy that the relay nodes should cover those nonreceiver and
algorithm will select a; b; c, and d as relay nodes. nonrelay nodes of the lower level.
We use a simple example to illustrate the process. There is
4.1.2 Multicast Structure a tree mesh in Fig. 5a, where nodes 6, 7, and 8 are
The broadcast structure mentioned above contains some multireceivers. First, we select node 4 at level 2 because it
unnecessary branches if the destinations do not involve all covers all the multireceivers at level 3. Next, we select node 2
the nodes. Instead, we propose to construct a “slim” at level 1, which covers all the multireceivers and the relay
structure by using the MCM Tree Construction algorithm node at level 2. Finally, we get the multicast tree in Fig. 5b.
described in Algorithm 3. 4.2 Channel Assignment
Algorithm 3. MCM tree construction algorithm The tree node discovery in the previous section allows each
Data: T : tree mesh of the network multireceiver to connect with the gateway through minimal
Result: T 0 : multicast tree hop count distance. In this section, we discuss how to assign
Use BFS to partition nodes into different levels; channels to the interfaces of the tree nodes by proposing
for 8 node v 2 V ðT Þ do two allocation algorithms: ascending channel allocation and
c[v] ¼ true if and only if v is a multi-receiver or the heuristic channel assignment.
source.
4.2.1 Ascending Channel Allocation
end
for l ¼ LevelNum  1; l >¼ 1; l ¼ l  1 do As assumed in Section 2, each node has two interfaces.
Si ¼ {node vi jvi belongs to level l  1}; Specially, the interface that a node uses to receive packets
from its relay node at the upper layer, termed Receive-
Sj ¼ {node vj jvj belongs to level l and c[vj ] ¼ true};
Interface (RI), is disjoint from the interface the node uses to
While Sj ! ¼ ; do
forward packets to its children, called Send-Interface (SI). In
Find vi1 ; vi2 ; . . . . in Sj with the minimal number of
order to guarantee that the relay node can communicate
parents;
with its children, each node’s RI is associated with the SI of
Among the parents of vi1 ; vi2 ; . . . ., find node tf with
its relay node, i.e., they should be assigned the same
the maximal number of children;
channel. Ascending Channel Allocation is proposed to
c[tf ] ¼ true;
assign channels and described in Algorithm 4.
Si ¼ Si  ftf g;
The children of tf record tf as their relay node; Algorithm 4. Ascending channel allocation algorithm
Sj ¼ Sj  fthe children of tf g; Data: f0; 1; . . . C  1g: available orthogonal channel set; T 0 :
end multicast tree;
end Result: Channel assignment for interfaces
V ðT 0 Þ ¼ ;; EðT 0 Þ ¼ ;; The source uses channel 0 for its SI;
for 8 node v 2 V ðT Þ do Its children use channel 0 for their RIs;
V ðT 0 Þ ¼ V ðT 0 Þ [ fvg if and only if c[v ] ¼ true; A ¼ 0;
edge e ¼ ðv; vs relay parent); for l ¼ 1; l  LevelNum-2; l++ do
EðT 0 Þ ¼ EðT 0 Þ [ feg; for 8 relay node u at level l do
end A ¼ ðA þ 1Þ mod C;
92 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO. 1, JANUARY 2010

area is approximated as a circle whose area is determined

by IR2 ðuv Þ. Since
X X
IR2 ðuv Þ ¼ ðR  jiu iv j Þ2 ;
v2NðuÞ v2NðuÞ
P 2
we just need to minimize v2NðuÞ jiu iv j . Based on this
consideration, we propose the Heuristic Channel Assign-
ment in Algorithm 5.
Fig. 6. Ascending channel allocation example. Algorithm 5. Heuristic channel assignment algorithm
Data: CH: available channel set; T 0 : multicast tree;
u uses channel A for its SI; Result: Channel assignment for interfaces
u’s children use channel A for their RIs; The source uses channel 0 for its SI;
end Its children use channel 0 for their RIs;
end for l ¼ 1; l  LevelNum-2; l++ do
The basic idea of the algorithm is straightforward: From for 8 relay node u at level l do
top to down in the tree, the channels are assigned to the S(u) ¼ {us neighboring relay nodes that have been
interfaces in the ascending order until the maximum channel assigned channels for their SIs}
P
number is reached, then start from channel 0 again. Although Choose channel i 2 CH that minimizes v2SðuÞ ji2 u iv j
simple, this approach avoids the situation that the same u uses channel i for its SI;
channel is assigned to two nearby links that interfere with u’s children use channel i for their RIs;
each other. We use a simple case to illustrate this algorithm in end
Fig. 6, where the number of the orthogonal channels are end
three. Note that the number above the node represents the
channel number used for its RI, while the number below
the node represents the channel number for its SI. 5 FURTHER DISCUSSION ON MCM ALGORITHM
In the algorithm, we only use limited orthogonal In this section, we discuss some companion mechanisms to
channels. 802:11b provides 11 channels in American domain further improve the MCM algorithm.
and 13 channels in European domain [8], 5 MHz apart in
frequency. However, to be totally orthogonal, the frequency 5.1 Repair Mechanism
should be at least 30 MHz, so 802:11b can offer only three Here, we discuss the failure recovery mechanism and node
nonoverlapping channels. Thus, although the Ascending join mechanism.
Channel Allocation is easy to implement, its performance is
still constrained by the limited number of orthogonal 5.1.1 Failure Recovery
channels. Fortunately, as mentioned in Section 2, network Usually the mesh routers work properly, but node failure
throughput can be further improved by exploiting all the can happen for various reasons. When a tree node fails,
partially overlapping channels. nodes in its subtree lose their connectivity to the root. Our
mechanism will reorganize the multicast tree to bypass the
4.2.2 Heuristic Channel Assignment
failed node and restore the connectivity.
In fact, we can utilize all the channels instead of just At first, we can safely assume that any node is able to
orthogonal channels. In Section 2, we observed that the
detect the failure of its neighbor quickly since the nodes
interference range decreases with the increase of the channel
periodically send “hello” messages to their neighbors. If a
separation. Intuitively, the channel assignment should make
node does not receive a hello message from one neighbor
a large channel separation for two wireless links if the
for a period of time, it considers the neighbor to have failed.
physical distance between them is short. We aim to minimize
There are two cases that apply to the node failure: the
the sum of the interference area of all the transmissions.
We use IRðuv Þ to indicate the interference range of collapsed node is a leaf or a relay node.
sender u of one link with respect to sender v of another For the first case, if the collapsed node v is a leaf, we
link. According to the experiment we performed in propose two approaches. One is just to leave it alone since
Section 2, under the condition that all the nodes have the leaf is not responsible for forwarding packets to any
the same transmission range R; IRðuv Þ ¼ R  jiu iv j . Here, other tree nodes. (The manner that the mesh clients within
u and v use channel iu and iv for their SIs, respectively, the communication range of the failed node restore the
and t is the Interference Factor. When allocating a connectivity to the network is beyond the scope of this
channel for relay node u, the channel assignment should paper.) This approach is simple, but the parent of the failure
take a channel that minimizes the sum of the square of node will continue to receive packets even if the parent is
the IRs between not a multireceiver.
P u and 2us neighboring relay nodes, that
is, minimize v2NðuÞ IR ðuv Þ, where NðuÞ represents the
The second choice is related with vs parent u on the tree.
set of the neighboring relay nodes of u. This is because If u has only one child on the tree, then it stops forwarding
the bigger interference area means the bigger chance two packets. Moreover, if u is not a multireceiver, it sends out a
transmissions may interfere. In addition, the interference message to its neighbors, announcing that it is no longer
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 93

Fig. 7. Repair mechanism example.

part of the tree. Then, the parent of u will do the same thing. 5.1.2 Node Join
The process continues until one ancestor of v has more than If a new node wants to join the multicast group, it sends out a
one child or it is a multireceiver. The first choice is simple to request for connecting to a nearby relay node. The request is
implement, and v is able to join the multicast group again if locally flooded, and the nodes on the path that reaches the
v recovers from failure after a short period of time. The nearest relay node will be absorbed into the multicast tree.
second choice helps to remove those unnecessary branches,
5.2 Channel Assignment with More Interfaces
which reduces the interference and saves bandwidth.
For the second case, when v is a relay node, all its The previous part of this paper assumes that each node has
children on the tree should check whether they are physical only two interfaces. We now discuss the case when there
neighbors of some other relay nodes on the tree. (If two are more interfaces for each node. In the multicast
application in WMNs, we can utilize multiple interfaces
nodes are within each other’s transmission range, they are
(more than three) to achieve parallel transmissions to
called physical neighbors even if they are not using the
further improve the throughput.
same channel.) If they are, the channels of their RIs will be
If each node has 2k interfaces, we divide the interfaces
reassigned as that of the “backup” relay node’s SI and into two groups: the sending group f1; 2; 3; . . . ; kg and the
reestablish the connectivity with the gateway. If they are not receiving group {k þ 1; k þ 2; . . . ; 2k}. The sending group is
physical neighbors of any relay node, each node will responsible for forwarding packets while the receiving
randomly choose one neighbor t at the upper layer, group is responsible for receiving packets.
requesting t to be its relay node. If t is the physical neighbor We still use the same multicast tree construction method
of one relay node, it connects with the relay node by using described in Section 4, but we modify the channel assign-
the same channel; otherwise, t will randomly choose one of ment algorithm. Suppose there are C available channels.
its neighbors at the upper layer, asking that node to be a First, we assume the node has only two interfaces: RI and SI,
relay node. This process continues until the request arrives and we use the channel assignment algorithms in Section 4
at a physical neighbor of any relay node. by restricting the number of available channels to Ck . After
We use an example in Fig. 7 to illustrate the repair the process, RI is assigned channel i and SI is assigned
mechanism. Figs. 7a and 7b give the network topology and channel j, respectively. We then apply this result to the
the responding multicast tree, respectively, where nodes 8, 2k interfaces: interface p (1  p  k) is assigned channel
9, 10, and 13 are multireceivers. If node 13 breaks down, i þ Cðp1Þ
k , and interface q (k þ 1  q  2k) is assigned
because it is a leaf, we can simply leave it alone. The other channel j þ Cðqk1Þ
k .
choice is that 13 requires its parent 7 to stop forwarding An example in Fig. 8 illustrates the mechanism. If C ¼ 6
packets. Node 7 realizes that it has one child on the tree, so and k ¼ 2, we allocate C 0 ¼ Ck ¼ 3 channels to the interfaces
it also asks its parent 3 to stop forwarding. The resulting as if each node has only two interfaces, and the result is
multicast tree after failure of node 13 is shown in Fig. 7c. shown in Fig. 8a. Next, we get the channel assignment for
If node 5 breaks down, its children begin to look for the 4-interface case, which is shown in Fig. 8b, where the
other connections to the source. Node 9 finds that it can numbers above the node represent the channels for the
communicate with relay node 4, so it changes the channel receiving group and the numbers below the node represent
on its RI for packet reception from 4. Node 10 cannot
communicate with any relay node on the tree, so it
randomly selects one of its physical neighbors at the upper
level, such as 6, requesting 6 to be its relay node. Node 6
then tries to communicate with any neighboring relay node,
and it sets the channel of its RI the same as the SI of relay
node 3. Node 6 then chooses a channel that is not used by
any of its neighbors for its SI, and node 10 accordingly sets
the same channel for its RI for packet reception. The
resulting multicast tree after the failure of node 5 is shown Fig. 8. Example of 4-interface channel allocation. (a) First step and
in Fig. 7d. (b) second step.
94 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO. 1, JANUARY 2010

Fig. 9. Example of multiple gateways. (a) Network topology, (b) multicast forest, (c) use gateway S1, and (d) use gateway S2.

the channels for the sending group. Now, we can see that 6 SIMULATIONS
node a can simultaneously transmit the packets to nodes b We evaluate the MCM algorithm by comparing it with the
and c on two wireless links, which are on channel 0 and 3,
LCA algorithm and a single channel multicast algorithm
respectively.
through the following metrics:
5.3 Multiple Gateways
. Throughput: The throughput is the average num-
Multiple gateways may be available as the Internet access ber of packets each multireceiver receives during a
points for a wireless mesh network in order to provide
time unit.
more bandwidth and improve the system throughput. In
. Delay: The delay is the average time it takes for a
the case of our multicast study with multiple gateways, the
packet to reach the destination after it leaves the
multireceivers can obtain the same information packets
source.
through different gateways. Therefore, one way to optimize
the multicast performance is to build multiple multicast We use an NS2 simulator (version 2.29) [10] to simulate a
trees initialized from different gateways to shorten the hop flat area of 900 m by 900 m with varying number of
count distance between each receiver to its root. We design randomly positioned wireless router nodes. By extending
the following algorithm to build multiple multicast trees the NS2 simulator, we configure all nodes to use multiple
rooted from different gateways. interfaces/channels with a transmission range of 250 m and
First, each gateway starts a breadth first search so that a carrier sensing range of 550 m. We use the default IEEE
each node in the network will get its own hop count
802.11 MAC configuration in NS2, which supports multi-
distances to each gateway.
casting using broadcasting at the base rate 1 Mbps.
Second, each multireceiver chooses the nearest gateway
We evaluate LCA and MCM algorithms in different
as its information root and notifies the gateway so that each
gateway will know which multireceivers belong to itself. scenarios. For each scenario, we randomly generate 100 dif-
Third, for each gateway constructed by using Algo- ferent graphs, where the source and the destinations are
rithm 3, a multicast tree rooted from this gateway is built randomly selected. Traffic is generated by constant bit rate
for the multireceivers that choose this gateway. (CBR) sessions. We vary the session rate at some scenarios.
We use a simple example to demonstrate the whole The packet size for all traffic is set to be 512 bytes. Except for
process. Fig. 9a shows a network topology, where nodes s1 the last section, we use the orthogonal channels in the
and s2 are two gateways and nodes a-e are the multi- simulation.
receivers. At the beginning, node s1 and node s2 initialize a
breadth first search, respectively. After the breadth first 6.1 Impact of Network Size
search, each multireceiver knows the hop count distances to We evaluate the throughput in different network sizes by
both s1 and s2 . Each multireceiver will select the gateway assigning the number of nodes with 30 and 60, and assigning
that is closer to it, so nodes a; b, and c choose s1 as their the number of the available channels with 12. We vary the
gateway, while nodes d and e choose s2 as their gateway. number of multireceivers from 5 to 25 in a 30-node-sized
Two multicast trees rooted from s1 and s2 will be built, network and from 5 to 55 in a 60-node-sized network. We
which is shown in Fig. 9b. measure the throughput of the MCM algorithm, the LCA
Instead, if we just use one gateway, the multicast tree algorithm, and the single-channel algorithm in which only
becomes larger. For example, if only gateway s1 is used, the one single channel is used in the multicasting. The results
multicast tree is built as shown in Fig. 9c, where the total
are shown in Fig. 10. We can see that using multichannel and
hop count distance is 12 and the number of relay nodes is 7.
multi-interface significantly improves the throughput. The
If only gateway s2 is used, the multicast tree is built as
reason is that using different channels prevents the channel
shown in Fig. 9d, where the total hop count distance is also
12 and the number of relay nodes is 7. On the other hand, if interference among close-by transmissions. Compared with
we use both s1 and s2 , the total hop count distance of the LCA, MCM further improves throughput, although they
multicast forest decreases to 9 and the number of relay both take advantage of multiple channels and multiple
nodes decreases to 4, so the multicast structure becomes interfaces. This is because MCM builds a more efficient
“slim” and the system throughput will be improved as multicast tree and carefully assigns the channels on the tree;
shown in next section. thus, it further reduces the interference.
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 95

Fig. 10. Impact of network size.

Fig. 11. Impact of Number of Channels.

6.2 Impact of Number of Channels the saturated rate almost does not help to improve the
We vary the number of available channels from 2 to 14 and throughput. Since MCM has the higher saturated transmis-
measure the throughput of MCM and LCA. Fig. 11 shows that sion rate, this means that MCM can take greater advantage of
MCM enhances throughput more than LCA when the the channel diversity than LCA.
number of channels is increased. In addition, we notice that
when the number of channels varies from 2 to 6, both MCM 6.4 Delay Comparison
and LCA have great throughput improvement. However, In this simulation, we evaluate the delay of LCM and MCM
when the number of channels varies from 7 to 14, MCM has by comparing the average time each packet takes to reach
small throughput improvement while LCA has almost no the destination. The transmission rate is set to 200 packet/s.
improvement. The explanation of the phenomenon on MCM Fig. 13 shows that MCM has a much shorter delay than
is that when the number of channels increases to a certain LCA. We also see that the delay of MCM decreases rapidly
extent, it is enough to eliminate almost all the interference in when the number of channels increases from 6 to 12, since
the network, thus using more channels cannot further interference is greatly reduced by using more channels.
improve throughput. At the same time, the number of 6.5 Two Interfaces versus Four Interfaces
channels that LCA uses is equal to the tree height; thus, some
In Section 5, we discussed how to allocate channels to more
channels are left unused when the number exceeds tree
interfaces. Now we evaluate the throughput of MCM on a
height, although some interference still exist in the network.
different number of interfaces by varying the channel
6.3 Impact of Transmission Rate numbers from 4 to 18, and the results are reported in Fig. 14.
We can see that using four interfaces usually can further
We vary the transmission rate from 50 packet/s to 300 pack-
improve the throughput, since utilizing more interfaces
et/s, and measure the throughput of MCM and LCA. Fig. 12
allows more simultaneous data transmissions. On the other
shows that MCM achieves much better throughput than LCA hand, making use of more interfaces also leads to more local
under different transmission rates. We also observe that the flow contentions. If there are not enough available channels,
saturated transmission rates for MCM and LCA to achieve the extra contentions caused by extra interfaces degrade the
nearly the maximal throughput are 225 packet/s and network performance. Thus, when the number of available
125 packet/s, respectively. The transmission rate exceeding channels is small (i.e., under six), using two interfaces is a
96 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO. 1, JANUARY 2010

Fig. 12. Impact of transmission rate.

Fig. 13. Delay comparison.

Fig. 14. Two Interfaces versus four interfaces.

better choice. Even if the number of channels exceeds six, As we know that the number of current available channels
the network throughput of four interfaces does not achieve is limited to under 14 and the number of orthogonal channels
nearly two times of that of two interfaces as expected until is much less, exploiting more interfaces does not achieve
the number reaches a threshold. Given the fixed network much better results subject to the current channel condition.
size, the threshold is related with the number of multi- In addition to the increased cost of equipping with more
receivers that decides the size of the multicast tree and the interfaces, using two interfaces currently seems to be a
contention level, that is, more receivers bring more local suitable choice. In the future, with exploiting more channels
contentions. The results show that the threshold for five and more low-cost interfaces, making use of more interfaces
receivers is 14, while it is 18 for 35 receivers. may have an important application.
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 97

Fig. 15. Partially overlapping channel test. (a) Spectrum Range: 30 MHz. (b) Spectrum Range: 60 MHz.

Fig. 16. Multiple gateway test.

6.6 Partially Overlapping Channel Assignment Test Allocation, since it makes a large channel separation for the
Usually, users are offered a range of frequency spectrum, adjacent wireless links.
where the numbers of orthogonal channels and partially
6.7 Multiple Gateways versus Single Gateway
overlapping channels are fixed. We choose two frequency
spectrums: 1) 30 MHz, which can offer two orthogonal There are usually multiple gateways in a wireless mesh
channels and six partially overlapping channels, and network to improve the throughput of the data flows between
2) 60 MHz, which can offer three orthogonal channels and the network and the Internet. In this section, we assume that
12 partially overlapping channels. In the two spectrums, we the wireless mesh network has two gateways. We compare
compare the throughput of the MCM algorithm under the the throughput of using both gateways with that of using only
following different channel assignment methods: one gateway under different network sizes of 30 nodes and
60 nodes. The number of the available channels is 12, and the
1. We only use the orthogonal channels defined in the number of multireceivers varies from 5 to 25 in a 30-node-
fixed range of frequency spectrum. sized network and from 5 to 55 in a 60-node-sized network.
2. We use Ascending Channel Allocation to allocate all The packet transmission interval is set as 0.018 sec/packet.
the partially overlapping channels in the frequency The results are shown in Fig. 16. We can see that using
spectrum to the interfaces. multiple gateways is able to significantly improve the
3. We use Heuristic Channel Assignment to allocate all network throughput because it can decrease the total hop
the partially overlapping channels in the frequency count distance of the multicast trees.
spectrum to the interfaces.
The results in Fig. 15 show that using partially over-
lapping channels can achieve better throughput than using
7 RELATED WORK
just orthogonal channels. This is because the orthogonal As a basic data communication mechanism, multicast has
channels are so scarce that they cannot eliminate all the been intensively studied in Internet and multihop wireless
interference, while the partially overlapping channels can networks. These proposed multicast algorithms and proto-
further reduce interference. We also observe that Heuristic cols do not take advantage of multichannels and multi-
Channel Assignment is better than Ascending Channel interfaces. At the same time, there exists a large number of
98 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. 21, NO. 1, JANUARY 2010

studies that aim to address the channel assignment problem present a novel multiple source active measurement
of mesh networks, which is characterized by using multi- procedure using a semirandomized probing scheme and
channels and multi-interfaces. Thus, we summarize multicast packet arrival order measurements. In [30], the author
research in MANET and representative research on multi- designs asymptotical algorithms for preprocessing to
channel in wireless networks. facilitate point-to-point distance queries under multiple
The multicast routing protocols in MANET are classified
source situation. A dynamic algorithm to maintain a
into three categories according to the way multicast routes are
minimum spanning forest in a planar embedded graph
created: 1) tree-based, 2) mesh-based, and 3) stateless multi-
when edge deletions and insertions frequently emerge has
cast. In the tree-based protocols, the data packets are
been proposed in [31].
transmitted from the source to the destinations along the
paths on the multicast tree, which helps to minimize the
bandwidth cost [11], [12]. The mesh-based protocols try to 8 CONCLUSION
create multiple trees among the group members such that the
In this paper, we investigate the multicast algorithm in
packets can be delivered to each receiver through multiple
wireless mesh networks where the throughput and the
paths [13], [14], [15]. The alternative paths help to increase the
delay have the paramount priorities. In order to achieve
protection against the topology alteration. However, both
efficient multicast in WMNs, two multicast algorithms,
tree-based and mesh-based multicast protocols have to bear
the overhead of creating and maintaining the multicast LCA and MCM, are proposed by using multichannels and
overhead keeping at the intermediate nodes. In order to multi-interfaces. An effective multicast structure is con-
address this drawback, stateless multicast protocols have structed to minimize the number of the relay nodes and the
been proposed to store the destination list in the packet communication delay. The dedicated channel assignment
header, and the packets are self-routed to the destinations helps to further reduce the interference as well. Compared
based on the geographical information [16], [17], [18]. with previous multicast approaches, our algorithms are
Our proposed MCM algorithm inherits the tree-based based on the multichannel and focus on the throughput
multicast design, but it differs from previous approaches in improvement. The performance evaluation shows that our
the following aspects: 1) Instead of power efficiency and algorithms outperform the single-channel multicast in
route recovery, which are usually the paramount goals of terms of throughput and delay, and more efficient multicast
multicast in MANET and sensor networks, MCM aims at structure and subtle channel assignment can further
maximizing throughput. 2) The multicast tree constructed improve throughput and reduce delay.
by MCM is formed by the diverse channels. Based on this,
MCM is able to minimize the interference by making use of
multichannels and multi-interfaces. ACKNOWLEDGMENTS
In recent years, multichannel issues in wireless networks The authors would like to thank the anonymous
have drawn much attention. Several link layer and MAC reviewers for their helpful comments. This work was
layer protocols have been proposed to improve the perfor- supported in part by the US National Science Foundation
mance of wireless networks [3], [19], [20], [21]. These under grants CCF-0514078, OCI-0753362, CNS-0551464,
approaches are to find the optimal channel for the current and CNS-0721441.
packet transmission for essentially avoiding interference.
Such schemes have the key advantage that a single radio is
required to support multiple channels. Some researchers aim REFERENCES
to derive the lower bound or upper bound of the capacity in [1] http://pdos.csail.mit.edu/roofnet/doku.php, 2009.
terms of achievable QoS in mesh networks [5], [22]. Many [2] http://www.seattlewireless.net, 2009.
studies focus on how to assign channels to nodes in the [3] J. So and N. Vaidya, “Multi-Channel Mac for Ad Hoc Networks:
Handling Multi-Channel Hidden Terminals Using a Single
network, either by static or dynamic methods [4], [23], [24],
Transceiver,” Proc. ACM MobiHoc, 2004.
[25], [26]. They develop a set of centralized or distributed [4] K. Ramachandran, E.M. Belding, K. Almeroth, and M. Buddhiko,
algorithms for channel assignments by taking the bandwidth “Interference-Aware Channel Assignment in Multi-Radio Wire-
cost, efficient routing, and load balance into account. Most of less Mesh Networks,” Proc. IEEE INFOCOM, 2006.
them believe that static assignment outperforms dynamic [5] J. Tang, G. Xue, and W. Zhang, “Maximum Throughput and Fair
Bandwidth Allocation in Multi-Channel Wireless Mesh Net-
assignment due to the channel switching cost and the delay. works,” Proc. IEEE INFOCOM, 2006.
New metrics have also been proposed for multihop wireless [6] A. Mishra, E. Rozner, S. Banerjee, and W. Arbaugh, “Exploiting
networks with considering the impact of channel inter- Partially Overlapping Channels in Wireless Networks: Turning a
ference, which is used to find high throughput paths Peril into an Advantage,” Proc. ACM/USENIX Internet Measure-
ment Conf., 2005.
between sources and destinations [27], [28]. [7] A. Mishra, V. Shrivastava, and S. Banerjee, “Partially Overlapped
Our channel assignment is different from previous Channels Not Considered Harmful,” Proc. ACM SIGMETRICS/
channel assignment approaches in two aspects: 1) we assign Performance, 2006.
channels based on the multicast structure so that we can [8] P. Li, N. Scalabrino, Y. Fang, E. Gregori, and I. Chlamtac,
“Channel Interference in IEEE 802.11b Systems,” Proc. IEEE Global
exploit the broadcast property of wireless nodes, and 2) we Telecomm. Conf. (GLOBECOM), 2007.
make full use of the partially overlapping channels instead [9] T.H. Cormen, C.E. Leiserson, R.L. Rivest, and C. Stein, Instroduc-
of just orthogonal channels, which ultimately reduces the tion to Algorithms. The MIT Press, 2001.
interference and improves the throughput. [10] http://www.isi.edu/nsnam/ns/index.html, 2009.
[11] E. Royer and C. Perkins, “Multicast Operation of the Ad-Hoc on
Previous research has been done on how to improve the Demand Distance Vector Routing Protocol,” Proc. ACM MobiCom,
network by utilizing multiple sources. Rabbat et al. [29] 1999.
ZENG ET AL.: EFFICIENT MULTICAST ALGORITHMS FOR MULTICHANNEL WIRELESS MESH NETWORKS 99

[12] J. Jetcheva and D.B. Johnson, “Adaptive Demand-Driven Multi- Guokai Zeng received the BS and MS degrees
cast Routing in Multi-Hop Wireless Ad Hoc Networks,” Proc. in computer science from University of Science
ACM MobiHoc, 2001. and Technology of China, in 2002 and 2005,
[13] S. Lee, M. Gerla, and C. Chiang, “On Demand Multicast Routing respectively. He is currently a PhD student in
Protocol,” Proc. IEEE Wireless Comm. and Networking Conf. (WCNC computer science at Michigan State University.
’99), pp. 1313-1317, Aug. 1999. His research interests are in the areas of
[14] J.J. Garcia-Luna-Aceves and E.L. Madruga, “The Core-Assisted distributed systems and wireless networking,
Mesh Protocol,” IEEE J. Selected Areas in Comm., vol. 17, no. 8, including wireless mesh networks, wireless sen-
pp. 1380-1394, Aug. 1999. sor networks, wireless ad hoc network, channel
[15] S. Das, B. Manoj, and C. Murthy, “A Dynamic Core Based assignment, QoS routing, and MAC layer design.
Multicast Routing Protocol,” Proc. ACM MobiHoc, 2002. He is a student member of the IEEE and the IEEE Computer Society.
[16] L. Ji and M.S. Corson, “Differential Destination Multicast-A Manet
Multicast Routing Protocol for Small Groups,” Proc. IEEE
INFOCOM ’01, pp. 1192-1201, Apr. 2001. Bo Wang is currently a PhD candidate in
[17] K. Chen and K. Nahrstedt, “Effective Location-Guided Overlay computer science at Michigan State University.
Multicast in Mobile Ad Hoc Networks,” Int’l J. Wireless and Mobile His research interests include wireless network-
Computing (IJWMC), vol. 3, 2005. ing and mobile computing. He is currently work-
[18] M. Mauve, H. Fuler, J. Widmer, and T. Lang, “Position-Based ing on resource allocation and QoS support in
Multicast Routing for Mobile Ad Hoc Networks,” Proc. ACM wireless mesh networks. He is a student member
MobiHoc ’03, June 2003. of the IEEE and the IEEE Computer Society.
[19] P. Bahl, R. Chandra, and J. Dunagan, “Ssch: Slotted Seeded
Channel Hopping for Capacity Improvement in IEEE 802.11 Ad-
Hoc Wireless Networks,” Proc. ACM MobiCom, 2004.
[20] Y. Liu and E. Knightly, “Opportunistic Fair Scheduling over
Multiple Wireless Channels,” Proc. IEEE INFOCOM, 2003.
[21] A. Tzamaloukas and J.J. Garcia-Luna-Aceves, “A Receiver
Initiated Collision-Avoidance Protocol for Multi-Channel Net- Yong Ding received the BS and MS degrees
works,” Proc. IEEE INFOCOM, 2001. from Southeast University, China, in 2001 and
[22] M. Kodialam and T. Nandagopal, “Characterizing the Capacity 2004, respectively. He is currently a PhD student
Region in Multi-Radio Multi-Channel Wireless Mesh Networks,” in computer science at Michigan State Univer-
Proc. ACM MobiCom, 2005. sity. His research interests are in the areas of
[23] B. Raman, “Channel Allocation in 802.11-Based Mesh Networks,” distributed systems and computer networking,
Proc. IEEE INFOCOM, 2006. including wireless sensor networks, vehicular ad
[24] A. Raniwala, K. Gopalan, and T.-c. Chiueh, “Centralized Channel hoc networks, and wireless mesh networks. He
Assignment and Routing Algorithms for Multi-Channel Wireless is a student member of the IEEE and the IEEE
Mesh Networks,” ACM Mobile Computing and Comm. Rev., vol. 8, Computer Society.
no. 2, pp. 50-65, 2004.
[25] A. Raniwala and T. cker Chiueh, “Architecture and Algorithms
for an IEEE 802.11-Based Multi-Channel Wireless Mesh Net-
work,” Proc. IEEE INFOCOM, 2005.
[26] M. Alicherry, R. Bhatia, and L. Li, “Joint Channel Assignment and Li Xiao received the BS and MS degrees in
Routing for Throughput Optimization in Multiradio Wireless computer science from Northwestern Polytech-
Mesh Networks,” Proc. ACM MobiCom, 2005. nic University, China, and the PhD degree in
[27] R. Draves, J. Padhye, and B. Zill, “Routing in Multi-Radio, Multi- computer science from the College of William
Hop Wireless Mesh Networks,” Proc. ACM MobiCom, 2004. and Mary in 2002. She is an associate professor
[28] S. Roy, D. Koutsonikolas, S. Das, and Y.C. Hu, “Highthroughput of computer science and engineering at Michi-
Multicast Routing Metrics in Wireless Mesh Networks,” Proc. 26th gan State University. Her research interests are
IEEE Int’l Conf. Distributed Computing Systems (ICDCS), 2006. in the areas of distributed and networking
[29] M. Rabbat, R. Nowak, and M. Coates, “Multiple Source, Multiple systems, overlay systems and applications,
Destination Network Tomography,” Proc. IEEE INFOCOM, 2004. and wireless sensor and mesh networks. She
[30] P.N. Klein, “Multiple-Source Shortest Paths in Planar Graphs,” is a member of the ACM, the IEEE, the IEEE Computer Society, and
Proc. 16th Ann. ACM-SIAM Symp. Discrete Algorithms, 2005. IEEE Women in Engineering.
[31] D. Eppstein, G.F. Italiano, R. Tamassia, R.E. Tarjan, J. Westbrook,
and M. Yung, “Maintenance of a Minimum Spanning Forest in a Matt W. Mutka received the BS degree in
Dynamic Plane Graph,” J. Algorithms, vol. 13, pp. 33-54, 1992. electrical engineering from the University of
Missouri-Rolla, the MS degree in electrical
engineering from Stanford University, and
the PhD degree in computer science from the
University of Wisconsin-Madison. He is on
the faculty of the Department of Computer
Science and Engineering, Michigan State Uni-
versity, East Lansing, where he is currently a
professor and a department chairperson. He
has been a visiting scholar at the University of Helsinki, Helsinki,
Finland, and a member of technical staff at Bell Laboratories in
Denver, Colorado. His current research interests include mobile
computing, wireless networking, and multimedia networking. He is a
senior member of the IEEE and the IEEE Computer Society.

. For more information on this or any other computing topic,

please visit our Digital Library at www.computer.org/publications/dlib.