A distributed three hop routing protocol to increase the capacity of hybrid wireless networks

A Distributed Three-Hop Routing Protocol
to Increase the Capacity of Hybrid
Wireless Networks
Haiying Shen, Senior Member, IEEE, Ze Li, and Chenxi Qiu
Abstract—Hybrid wireless networks combining the advantages of both mobile ad-hoc networks and infrastructure wireless networks
have been receiving increased attention due to their ultra-high performance. An efficient data routing protocol is important in such
networks for high network capacity and scalability. However, most routing protocols for these networks simply combine the ad-hoc
transmission mode with the cellular transmission mode, which inherits the drawbacks of ad-hoc transmission. This paper presents a
Distributed Three-hop Routing protocol (DTR) for hybrid wireless networks. To take full advantage of the widespread base stations,
DTR divides a message data stream into segments and transmits the segments in a distributed manner. It makes full spatial reuse of a
system via its high speed ad-hoc interface and alleviates mobile gateway congestion via its cellular interface. Furthermore, sending
segments to a number of base stations simultaneously increases throughput and makes full use of widespread base stations. In
addition, DTR significantly reduces overhead due to short path lengths and the elimination of route discovery and maintenance. DTR
also has a congestion control algorithm to avoid overloading base stations. Theoretical analysis and simulation results show the
superiority of DTR in comparison with other routing protocols in terms of throughput capacity, scalability, and mobility resilience. The
results also show the effectiveness of the congestion control algorithm in balancing the load between base stations.
Index Terms—Hybrid wireless networks, routing algorithm, load balancing, congestion control
Ç
1 INTRODUCTION
OVER the past few years, wireless networks including
infrastructure wireless networks and mobile ad-hoc
networks (MANETs) have attracted significant research
interest. The growing desire to increase wireless network
capacity for high performance applications has stimulated
the development of hybrid wireless networks [1], [2], [3],
[4], [5], [6]. A hybrid wireless network consists of both an
infrastructure wireless network and a mobile ad-hoc net-
work. Wireless devices such as smart-phones, tablets and
laptops, have both an infrastructure interface and an ad-hoc
interface. As the number of such devices has been increas-
ing sharply in recent years, a hybrid transmission structure
will be widely used in the near future. Such a structure syn-
ergistically combines the inherent advantages and over-
come the disadvantages of the infrastructure wireless
networks and mobile ad-hoc networks.
In a mobile ad-hoc network, with the absence of a central
control infrastructure, data is routed to its destination
through the intermediate nodes in a multi-hop manner. The
multi-hop routing needs on-demand route discovery or
route maintenance [7], [8], [9], [10]. Since the messages are
transmitted in wireless channels and through dynamic rout-
ing paths, mobile ad-hoc networks are not as reliable as
infrastructure wireless networks. Furthermore, because of
the multi-hop transmission feature, mobile ad-hoc networks
are only suitable for local area data transmission.
The infrastructure wireless network (e.g., cellular net-
work) is the major means of wireless communication in our
daily lives. It excels at inter-cell communication (i.e., com-
munication between nodes in different cells) and Internet
access. It makes possible the support of universal network
connectivity and ubiquitous computing by integrating all
kinds of wireless devices into the network. In an infrastruc-
ture network, nodes communicate with each other through
base stations (BSes). Because of the long distance one-hop
transmission between BSes and mobile nodes, the infra-
structure wireless networks can provide higher message
transmission reliability and channel access efficiency, but
suffer from higher power consumption on mobile nodes
and the single point of failure problem [11].
A hybrid wireless network synergistically combines an
infrastructure wireless network and a mobile ad-hoc net-
work to leverage their advantages and overcome their
shortcomings, and finally increases the throughput capacity
of a wide-area wireless network. A routing protocol is a crit-
ical component that affects the throughput capacity of a
wireless network in data transmission. Most current routing
protocols in hybrid wireless networks [1], [5], [6], [12], [13],
[14], [15], [16], [17], [18] simply combine the cellular trans-
mission mode (i.e., BS transmission mode) in infrastructure
wireless networks and the ad-hoc transmission mode in
mobile ad-hoc networks [7], [8], [9]. That is, as shown in
Fig. 1a, the protocols use the multi-hop routing to forward a
message to the mobile gateway nodes that are closest to the
BSes or have the highest bandwidth to the BSes. The band-
width of a channel is the maximum throughput (i.e.,
The authors are with the Department of Electrical and Computer
Engineering, Clemson University, Clemson, SC 29634.
E-mail: {shenh, zel, chenxiq}@clemson.edu.
Manuscript received 18 Mar. 2014; accepted 18 Dec. 2014. Date of publication
7 Jan. 2015; date of current version 31 Aug. 2015.
For information on obtaining reprints of this article, please send e-mail to:
reprints@ieee.org, and reference the Digital Object Identifier below.
Digital Object Identifier no. 10.1109/TMC.2015.2388476
IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 14, NO. 10, OCTOBER 2015 1975
1536-1233 ß 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
For More Details Contact G.Venkat Rao
8143271457
8143271457

transmission rate in bits/s) that can be achieved. The mobile
gateway nodes then forward the messages to the BSes, func-
tioning as bridges to connect the ad-hoc network and the
infrastructure network.
However, direct combination of the two transmission
modes inherits the following problems that are rooted in
the ad-hoc transmission mode.
High overhead. Route discovery and maintenance incur
high overhead. The wireless random access medium
access control (MAC) required in mobile ad-hoc net-
works, which utilizes control handshaking and a
back-off mechanism, further increases overhead.
Hot spots. The mobile gateway nodes can easily
become hot spots. The RTS-CTS random access, in
which most traffic goes through the same gateway,
and the flooding employed in mobile ad-hoc rout-
ing to discover routes may exacerbate the hot spot
problem. In addition, mobile nodes only use the
channel resources in their route direction, which
may generate hot spots while leave resources in
other directions under-utilized. Hot spots lead to
low transmission rates, severe network congestion,
and high data dropping rates.
Low reliability. Dynamic and long routing paths lead
to unreliable routing. Noise interference and neigh-
bor interference during the multi-hop transmission
process cause a high data drop rate. Long routing
paths increase the probability of the occurrence of
path breakdown due to the highly dynamic nature
of wireless ad-hoc networks.
These problems become an obstacle in achieving high
throughput capacity and scalability in hybrid wireless net-
works. Considering the widespread BSes, the mobile nodes
have a high probability of encountering a BS while moving.
Taking advantage of this feature, we propose a Distributed
Three-hop Data Routing protocol (DTR). In DTR, as shown
in Fig. 1b, a source node divides a message stream into a
number of segments. Each segment is sent to a neighbor
mobile node. Based on the QoS requirement, these mobile
relay nodes choose between direct transmission or relay
transmission to the BS. In relay transmission, a segment is
forwarded to another mobile node with higher capacity to a
BS than the current node. In direct transmission, a segment
is directly forwarded to a BS. In the infrastructure, the seg-
ments are rearranged in their original order and sent to the
destination. The number of routing hops in DTR is confined
to three, including at most two hops in the ad-hoc transmis-
sion mode and one hop in the cellular transmission mode.
To overcome the aforementioned shortcomings, DTR tries
to limit the number of hops. The first hop forwarding dis-
tributes the segments of a message in different directions to
fully utilize the resources, and the possible second hop for-
warding ensures the high capacity of the forwarder. DTR
also has a congestion control algorithm to balance the traffic
load between the nearby BSes in order to avoid traffic con-
gestion at BSes.
Using self-adaptive and distributed routing with high-
speed and short-path ad-hoc transmission, DTR signifi-
cantly increases the throughput capacity and scalability of
hybrid wireless networks by overcoming the three short-
comings of the previous routing algorithms. It has the fol-
lowing features:
Low overhead. It eliminates overhead caused by route
discovery and maintenance in the ad-hoc transmis-
sion mode, especially in a dynamic environment.
Hot spot reduction. It alleviates traffic congestion at
mobile gateway nodes while makes full use of chan-
nel resources through a distributed multi-path relay.
High reliability. Because of its small hop path length
with a short physical distance in each step, it allevi-
ates noise and neighbor interference and avoids the
adverse effect of route breakdown during data trans-
mission. Thus, it reduces the packet drop rate and
makes full use of spacial reuse, in which several
source and destination nodes can communicate
simultaneously without interference.
The rest of this paper is organized as follows. Section 2
presents a review of representative hybrid wireless net-
works and multi-hop routing protocols. Section 3 details the
DTR protocol, with an emphasis on its routing methods,
segment structure, and BS congestion control. Section 4 the-
oretically analyzes the performance of the DTR protocol.
Section 5 shows the performance of the DTR protocol in
comparison to other routing protocols. Finally, Section 6
concludes the paper.
2 RELATED WORK
In order to increase the capacity of hybrid wireless net-
works, various routing methods with different features
have been proposed. One group of routing methods inte-
grate the ad-hoc transmission mode and the cellular trans-
mission mode [1], [5], [6], [14], [16], [17], [18]. Dousse et al.
[6] built a Poisson Boolean model to study how a BS
increases the capacity of a MANET. Lin and Hsu [5] pro-
posed a Multihop Cellular Network (MCN) and derived its
throughput. Hsieh and Sivakumar [14] investigated a
hybrid IEEE 802.11 network architecture with both a distrib-
uted coordination function and a point coordination func-
tion. Luo et al. [1] proposed a unified cellular and ad-hoc
network architecture for wireless communication. Cho and
Haas [16] studied the impact of concurrent transmission in
a downlink direction (i.e., from BSes to mobile nodes) on
the system capacity of a hybrid wireless network. In [17],
[18], a node initially communicates with other nodes using
an ad-hoc transmission mode, and switches to a cellular
Fig. 1. Traditional and proposed routing algorithms on the uplink
direction.
1976 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 14, NO. 10, OCTOBER 2015
8143271457
8143271457

transmission mode when its performance is better than the
ad-hoc transmission.
The above methods are only used to assist intra-cell ad-
hoc transmission rather than inter-cell transmission. In
inter-cell transmission [1], [5], [6], a message is forwarded
via the ad-hoc interface to the gateway mobile node that is
closest to or has the highest uplink transmission bandwidth
to a BS. The gateway mobile node then forwards the mes-
sage to the BS using the cellular interface. However, most of
these routing protocols simply combine routing schemes in
ad-hoc networks and infrastructure networks, hence inherit
the drawbacks of the ad-hoc transmission mode as
explained previously.
DTR is similar to the Two-hop transmission protocol [19]
in terms of the elimination of route maintenance and the
limited number of hops in routing. In Two-hop, when a
node’s bandwidth to a BS is larger than that of each neigh-
bor, it directly sends a message to the BS. Otherwise, it
chooses a neighbor with a higher channel and sends a mes-
sage to it, which further forwards the message to the BS.
DTR is different from Two-hop in three aspects. First, Two-
hop only considers the node transmission within a single
cell, while DTR can also deal with inter-cell transmission,
which is more challenging and more common than intra-
cell communication in the real world. Second, DTR uses dis-
tributed transmission involving multiple cells, which makes
full use of system resources and dynamically balances the
traffic load between neighboring cells. In contrast, Two-hop
employs single-path transmission.
There are other methods proposed to improve routing
performance in hybrid wireless networks. Wu et al. [3]
proposed using ad-hoc relay stations to dynamically relay
traffic from one cell to another in order to avoid traffic con-
gestion in BSes. Li et al. [20] surveyed a number of multi-
hop cellular network architectures in literature, and
compared and discussed methods to reduce the cost of
deployment for MCNs. The work in [21] investigates how
to allocate the bandwidth to users to improve the perfor-
mance of hybrid wireless networks. Thulasiraman and
Shen [22] further considered the wireless interference in
optimizing the resource allocation in hybrid wireless net-
works. The work in [23] proposes a coalitional game the-
ory based cooperative packet delivery scheme in hybrid
wireless networks. There are also some works [24], [25],
[26] that study radio frequency allocation for direction
transmission and relay transmission in hybrid wireless
networks. These works are orthogonal to our study in this
paper and can be incorporated into DTR to further
enhance its performance.
The throughput capacity of the hybrid wireless net-
work under different settings has also been an active
research topic in the hybrid wireless network. The works
in [17], [27] have studied the throughput of hybrid net-
work with n nodes and m stations. Liu et al. [28] theoreti-
cally studied the capacity of hybrid wireless networks
under an one-dimensional network topology and a two-
dimensional strip topology. Wang et al. [29] studied the
multicast throughput of hybrid wireless networks and
designed an optimal multicast strategy based on deduced
throughput.
3 DISTRIBUTED THREE-HOP ROUTING PROTOCOL
3.1 Assumption and Overview
Since BSes are connected with a wired backbone, we assume
that there are no bandwidth and power constraints on trans-
missions between BSes. We use intermediate nodes to denote
relay nodes that function as gateways connecting an infra-
structure wireless network and a mobile ad-hoc network.
We assume every mobile node is dual-mode; that is, it has
ad-hoc network interface such as a WLAN radio interface
and infrastructure network interface such as a 3G cellular
interface.
DTR aims to shift the routing burden from the ad-hoc
network to the infrastructure network by taking advantage
of widespread base stations in a hybrid wireless network.
Rather than using one multi-hop path to forward a message
to one BS, DTR uses at most two hops to relay the segments
of a message to different BSes in a distributed manner, and
relies on BSes to combine the segments. Fig. 2 demonstrates
the process of DTR in a hybrid wireless network. We sim-
plify the routings in the infrastructure network for clarity.
As shown in the figure, when a source node wants to trans-
mit a message stream to a destination node, it divides the
message stream into a number of partial streams called seg-
ments and transmits each segment to a neighbor node.
Upon receiving a segment from the source node, a neigh-
bor node locally decides between direct transmission and
relay transmission based on the QoS requirement of the
application. The neighbor nodes forward these segments in
a distributed manner to nearby BSes. Relying on the infra-
structure network routing, the BSes further transmit the
segments to the BS where the destination node resides. The
final BS rearranges the segments into the original order and
forwards the segments to the destination. It uses the cellu-
lar IP transmission method [30] to send segments to the
destination if the destination moves to another BS during
segment transmission.
Our DTR algorithm avoids the shortcomings of ad-hoc
transmission in the previous routing algorithms that
directly combine an ad-hoc transmission mode and a cellu-
lar transmission mode. Rather than using the multi-hop ad-
hoc transmission, DTR uses two hop forwarding by relying
on node movement and widespread base stations. All other
aspects remain the same as those in the previous routing
algorithms (including the interaction with the TCP layer).
DTR works on the Internet layer. It receives packets from
Fig. 2. Data transmission in the DTR protocol.
SHEN ET AL.: A DISTRIBUTED THREE-HOP ROUTING PROTOCOL TO INCREASE THE CAPACITY OF HYBRID WIRELESS NETWORKS 1977
8143271457
8143271457

the TCP layer and routes it to the destination node, where
DTR forwards the packet to the TCP layer.
The data routing process in DTR can be divided into
two steps: uplink from a source node to the first BS and
downlink from the final BS to the data’s destination. Criti-
cal problems that need to be solved include how a source
node or relay node chooses nodes for efficient segment
forwarding, and how to ensure that the final BS sends
segments in the right order so that a destination node
receives the correct data. Also, since traffic is not evenly
distributed in the network, how to avoid overloading
BSes is another problem. Below, Section 3.2 will present
the details for forwarding node selection in uplink trans-
mission and Section 3.3 will present the segment structure
that helps ensure the correct final order of segments in a
message, and DTR’s strategy for downlink transmission.
Section 3.4 will present the congestion control algorithm
for balancing a load between BSes.
3.2 Uplink Data Routing
A long routing path will lead to high overhead, hot spots
and low reliability. Thus, DTR tries to limit the path length.
It uses one hop to forward the segments of a message in a
distributed manner and uses another hop to find high-
capacity forwarder for high performance routing. As a
result, DTR limits the path length of uplink routing to two
hops in order to avoid the problems of long-path multi-hop
routing in the ad-hoc networks. Specifically, in the uplink
routing, a source node initially divides its message stream
into a number of segments, then transmits the segments to
its neighbor nodes. The neighbor nodes forward segments
to BSes, which will forward the segments to the BS where
the destination resides.Below, we first explain how to define
capacity, then introduce the way for a node to collect the
capacity information from its neighbors, and finally present
the details of the DTR routing algorithm.
Different applications may have different QoS require-
ments, such as efficiency, throughput, and routing speed.
For example, delay-tolerant applications (e.g., voice mail,
e-mail and text messaging) do not necessarily need fast
real-time transmission and may make throughput the
highest consideration to ensure successful data transmis-
sion. Some applications may take high mobility as their
priority to avoid hot spots and blank spots. Hot spots are
areas where BS channels are congested, while blank spots
are areas without signals or with very weak signals. High-
mobility nodes can quickly move out of a hot spot or
blank spot and enter a cell with high bandwidth to a BS,
thus providing efficient data transmission. Throughput
can be measured by bandwidth, mobility can be measured
by the speed of node movement, and routing speed can be
measured by the speed of data forwarding. Bandwidth
can be estimated using the non-intrusive technique pro-
posed in [31]. In this work, we take throughput and rout-
ing speed as examples for the QoS requirement. We use a
bandwidth/queue metric to reflect node capacity in through-
put and fast data forwarding. The metric is the ratio of a
node’s channel bandwidth to its message queue size. A
larger bandwidth/queue value means higher throughput
and message forwarding speed, and vice versa.
When choosing neighbors for data forwarding, a node
needs the capacity information (i.e., queue size and band-
width) of its neighbors. Also, a selected neighbor should
have enough storage space for a segment. To keep track of
the capacity and storage space of its neighbors, each node
periodically exchanges its current capacity and storage
information with its neighbors. In the ad-hoc network com-
ponent, every node needs to periodically send “hello” mes-
sages to identify its neighbors. Taking advantage of this
policy, nodes piggyback the capacity and storage informa-
tion onto the “hello” messages in order to reduce the over-
head caused by the information exchanges. If a node’s
capacity and storage space are changed after its last “hello”
message sending when it receives a segment, it sends its
current capacity and storage information to the segment for-
warder. Then, the segment forwarder will choose the high-
est capacity nodes in its neighbors based on the most
updated information.
When a source node sends out message segments, it
chooses the neighbors that have enough space for storing a
segment, and then chooses neighbors that have the highest
capacity. In order to find higher capacity forwarders in a
larger neighborhood around the source, each segment
receiver further forwards its received segment to its neigh-
bor with the highest capacity. That is, after a neighbor node
mi receives a segment from the source, it uses either direct
transmission or relay transmission. If the capacity of each of
its neighbors is no greater than itself, relay node mi uses
direct transmission. Otherwise, it uses relay transmission.
In direct transmission, the relay node sends the segment to
a BS if it is in a BS’s region. Otherwise, it stores the segment
while moving until it enters a BS’s region. In relay transmis-
sion, relay node mi chooses its highest-capacity neighbor as
the second relay node based on the QoS requirement. The
second relay node will use direct transmission to forward
the segment directly to a BS. As a result, the number of
transmission hops in the ad-hoc network component is con-
fined to no more than two. The small number of hops help
to increase the capacity of the network and reduce channel
contention in ad-hoc transmission. Algorithm 1 shows the
pseudo-code for neighbor node selection and message for-
warding in DTR.
The purpose of the second hop selection is to find a
higher capacity node as the message forwarder in order to
improve the performance of the QoS requirement. As the
neighborhood scope of a node for high-capacity node
searching grows, the probability of finding higher capacity
nodes increases. Thus, a source node’s neighbors are more
likely to find neighbors with higher capacities than the
source node. Therefore, transmitting data segments to
neighbors and enabling them to choose the second relays
help to find higher capacity nodes to forward data. If a
source node has the highest capacity in its region, the seg-
ments will be forwarded back to the source node according
to the DTR protocol. The source node then forwards the seg-
ments to the BSes directly due to the three-hop limit.
Though sending data back and forth leads to latency and
bandwidth wastage, this case occurs only when the source
nodes is the highest capacity node within its two-hop neigh-
borhood. Also, this step is necessary for finding the highest
capacity nodes within the source’s two-hop neighborhood,
8143271457
8143271457

and ensures that the highest capacity nodes are always
selected as the message forwarders. If the source node does
not distribute segments to its neighbors, the higher-capacity
node searching cannot be conducted. Note that the data
transmission rate of the ad-hoc interface (e.g., IEEE 802.11)
is more than 10 times faster than the cellular interface (e.g.,
GSM, 3G). Thus, the transmission delay for sending the
data back and forth in the ad-hoc transmission is negligible
in the total routing latency.
Algorithm 1. Pseudo-Code for Neighbor Node Selection
and Message Forwarding.
1: ChooseRelay( ) f
2: //choose neighbors with sufficient caches and bandwidth/queue
(b/q) rates
3: Query storage size and QoS requirement info. from neighbors
4: for each neighbor n do
5: if n.cache.size segment.length n.b/q this.b/q then
6: Add n to R ¼ fr1; . . . rm; . . .g in a descending order of
b/q
7: end if
8: end for
9: Return R
10: g
11: Transmission( ) f
12: if it is a source node then
13: //routing conducted by a source node
14: //choose relay nodes based on QoS requirement
15: R ¼ ChooseRelay( );
16: Send segments to fr1; . . . rmg in R
17: else
18: //routing conducted by a neighbor node
19: if this.b/q b/q of all neighbors then
20: //direct transmission
21: if within the range of a BS then
22: Transmit the segment directly to the BS
23: end if
24: else
25: //relay transmission
26: nodei ¼ getHighestCapability(ChooseRelay( ))
27: Send a segment to nodei
28: end if
29: end if
30: g
By distributing a message’s segments to different nodes
to be forwarded in different directions, our algorithm
reduces the congestion in the previous routing algorithms
in the hybrid wireless networks. When a node selects a relay
to forward a segment, it checks the capacity of the node.
Only when a node, say node mi, has enough capacity, the
node will forward a segment to node mi. Therefore, even
though the paths are not node-disjoint, there will be no con-
gestion in the common sub-paths.
Fig. 3 shows examples of neighbor selection in DTR, in
which the source node is in the transmission range of a BS.
In the figures, the value in the node represents its capacity.
In scenario (a), there exist nodes that have higher capacity
than the source node within the source’s two-hop neighbor-
hood. If a routing algorithm directly let a source node trans-
mit a message to its BS, the high routing performance
cannot be guaranteed since the source node may have very
low capacity. In DTR, the source node sends segments to its
neighbors, which further forward the segments to nodes
with higher capacities. In scenario (b), the source node has
the highest capacity among the nodes in its two-hop neigh-
borhood. After receiving segments from the source node,
some neighbors forward the segments back to the source
node, which sends the message to its BS. Thus, DTR always
arranges data to be forwarded by nodes with high capacity
to their BSes. DTR achieves higher throughput and faster
data forwarding speed by taking into account node capacity
in data forwarding.
3.3 Downlink Data Routing and Data
Reconstruction
As mentioned above, the message stream of a source node
is divided into several segments. After a BS receives a seg-
ment, it needs to forward the segment to the BS, where the
destination node resides (i.e., the destination BS). We use
the mobile IP protocol [32] to enable BSes to know the
destination BS. In this protocol, each mobile node is asso-
ciated with a home BS, which is the BS in the node’s home
network, regardless of its current location in the network.
The home network of a node contains its registration
information identified by its home address, which is a
static IP address assigned by an ISP. In a hybrid wireless
network, each BS periodically emits beacon signals to
locate the mobile nodes in its range. When a mobile node
mi moves away from its home BS, the BS where mi cur-
rently resides detects mi and sends its IP address to the
home BS of mi. When a BS wants to contact mi, it contacts
the home BS of mi to find the destination BS where mi cur-
rently resides at.
However, the destination BS recorded in the home BS
may not be the most up-to-date destination BS since destina-
tion mobile nodes switch between the coverage regions of
different BSes during data transmission to them. For
instance, data is transmitted to BS Bi that has the data’s des-
tination, but the destination has moved to the range of BS
Bj before the data arrives at BS Bi. To deal with this prob-
lem, we adopt the Cellular IP protocol [30] for tracking
node locations. With this protocol, a BS has a home agent
and a foreign agent. The foreign agent keeps track of mobile
nodes moving into the ranges of other BSes. The home agent
intercepts in-coming segments, reconstructs the original
data, and re-routes it to the foreign agent, which then for-
wards the data to the destination mobile node.
Fig. 3. Neighbor selection in DTR.
8143271457
8143271457

After the destination BS receives the segments of a mes-
sage, it rearranges the segments into the original message
and then sends it to the destination mobile node. A vital
issue is guaranteeing that the segments are combined in the
correct order. For this purpose, DTR specifies the segment
structure format. Each segment contains eight fields, includ-
ing: (1) source node IP address (denoted by S); (2) destina-
tion node IP address (denoted by D); (3) message sequence
number (denoted by m); (4) segment sequence number
(denoted by s); (5) QoS indication number (denoted by q);
(6) data; (7) length of the data; and (8) checksum. Fields (1)-
(5) are in the segment head.
The role of the source IP address field is to inform the des-
tination node where the message comes from. The
destination IP address field indicates the destination node,
and is used to locate the final BS. After sending out a mes-
sage stream to a destination, a source node may send out
another message stream to the same destination node. The
message sequence number differentiates the different message
streams initiated by the same source node. The segment
sequence number is used to find the correct transmission
sequence of the segments for transmission to a destination
node. The data is the actual information that a source node
wants to transmit to a destination node. The length field
specifies the length of the DTR segment including the
header in bytes. The checksum is used by the receiver node
to check whether the received data has errors. The QoS indi-
cation number is used to indicate the QoS requirement of the
application.
Thus, each segment’s head includes the information
represented by ðS; D; m; s; qÞðm; s ¼ 1; 2; 3; . . .Þ. When a
segment with head ðS; D; m; s; qÞ arrives at a BS, the BS
contacts D’s home BS to find the destination BS where D
stays via the mobile IP protocol. It then transmits the seg-
ment to the destination BS through the infrastructure net-
work component. After arriving at the BS, the segment
waits in the cache for its turn to be transmitted to its des-
tination node based on its message and segment sequence
numbers. At this time, if another segment comes with a
head labelled ðS; D; ðm þ 1Þ; s; qÞ, which means that it is
from the same source node but belongs to another data
stream, the BS will put it to another stream. If the segment
is labeled as ðS; D; m; ðs þ 1Þ; qÞ, it means that this seg-
ment belongs to the same data stream of the same source
node as segment ðS; D; m; s; qÞ. The combination of the
source node’s sequence number and segment sequence
number helps to locate the stream and the position of a
segment in the steam. In order to integrate the segments
into their correct order to retrieve the original data, the
segments in the BS are transmitted to the destination
node in the order of the segments’ sequence in the origi-
nal message. If a segment has not arrived at the final BS,
its subsequent segments will wait in the final BS until its
arrival. Algorithm 2 shows the pseudo-code for a BS to
reorder and forward segments to their destinations. Note
that in the cache, we can set the timer based on the packet
rate and storage limit. In other words, the timer should be
set as large as possible to fully utilize the storage on BSes
to ensure that a message has a high probability to be
recovered.
Algorithm 2. Pseudo-Code for a BS to Reorder and
Forward Segments to Destination Nodes.
1: //a cache pool is built for each data stream
2: //there are n cache pools currently
3: if receives a segment (S, D, m, s, q) then
4: if there is no cache pool with msg sequence num equals
m then
5: Create a cache pool n þ 1 for the stream m
6: else
7: //the last delivered segment of stream m has sequence num
i À 1
8: if s ¼¼ i then
9: Send out segment (S, D, m, s, q) to D
10: i þ þ;
11: else
12: Add segment (S, D, m, s) into cache pool m
13: end if
14: end if
15: end if
3.4 Congestion Control in Base Stations
Compared to the previous routing algorithms in hybrid
wireless networks, DTR can distribute traffic load among
mobile nodes more evenly. Though the distributed routing
in DTR can distribute traffic load among nearby BSes, if the
traffic load is not distributed evenly in the network, some
BSes may become overloaded while other BSes remain
lightly loaded. We propose a congestion control algorithm
to avoid overloading BSes in uplink transmission (e.g., B1,
B2 and B3 in Fig. 1b) and downlink transmission (e.g., B4 in
Fig. 1b), respectively.
In the hybrid wireless network, BSes send beacon mes-
sages to identify nearby mobile nodes. Taking advantage of
this beacon strategy, once the workload of a BS, say Bi,
exceeds a pre-defined threshold, Bi adds an extra bit in its
beacon message to broadcast to all the nodes in its transmis-
sion range. Then, nodes near Bi know that Bi is overloaded
and will not forward segments to Bi. When a node near Bi,
say mi, needs to forward a segment to a BS, it will send the
segment to Bi based on the DTR algorithm. In our conges-
tion control algorithm, because Bi is overloaded, rather
than targeting Bi, mi will forward the segment to a lightly
loaded neighboring BS of Bi. To this end, node mi first
queries a multi-hop path to a lightly loaded neighboring BS
of Bi. Node mi broadcasts a query message into the system.
We set the TTL for the path query forwarding step to a con-
stant (e.g., 3). The query message is forwarded along other
nodes until a node (say mj) near a lightly loaded BS (say Bj)
is reached. Due to broadcasting, a node may receive multi-
ple copies of the same queries. Each node only remembers
mi and the node that forwards the first query (i.e., its pre-
ceding node), and ignores all other the same queries. In this
way, a multi-hop path between the source node and the
lightly loaded base station can be formed. Node
mj responds to the path query by adding a reply bit and the
address of mi into its beacon message to its preceding node
in the path. This beacon receiver also adds a reply bit and
the address of mi into its beacon message to its preceding
node in the path. This process repeats until mi receives the
beacon. Thus, each node knows its preceding node and
8143271457
8143271457

succeeding node in the path from mi and mj based on the
address of mi. Then, mi’s message can be forwarded along
the observed path along the nodes. The observed path can
always be used by mi for any subsequent messages to Bj as
long as it is not broken. The neighboring BSes of an over-
loaded BS may also be overloaded. As the mobile nodes
near an overloaded BS know that the BS is overloaded,
when they receive a query message to find a path to an
underloaded BS, they do not forward the message towards
their overloaded BSes.
Node mi may receive responses from a few nodes near
BSes. It can choose bðb ! 1Þ neighboring BSes of the destina-
tion to forward the segment. The redundant transmissions
enhance the data transmission reliability while also increase
the routing overhead. Thus, the value of b should be care-
fully determined based on the available resources for rout-
ing and the reliability demand. If b is set to a large value, the
routing reliability is high at the cost of high overhead. If b is
set to a small value, the routing reliability is low while the
overhead is reduced. After the neighboring BSes receive the
segments, they further forward the segments to the destina-
tion BS, which forwards the segments to the destination
node. In this way, the heavy traffic from mobile nodes to a
BS can be distributed among neighboring BSes quickly.
Next, we discuss how to handle the case when the desti-
nation BS is congested. If a BS has not received confirma-
tion from the destination BS during a certain time period
after it sends out a segment, it assumes that the destination
BS is overloaded. Then, it sends the segment to b ðb ! 1Þ
lightly loaded neighboring BSes of the destination BS from
its routing table. If an attempted neighboring BS does not
respond during a certain time period, it is also considered
as overloaded. Then, the BS keeps trying other neighboring
BSes until finding lightly loaded BSes. Redundant neigh-
boring BSes are selected in order to increase routing reli-
ability. The constant b should be set to an appropriate
value considering factors such as the network size and the
amount of traffic in order to achieve an optimal trade-off
between overhead and reliability.
After receiving the message, each lightly loaded neigh-
boring BS of the destination BS finds a multi-hop path to the
destination mobile node. It broadcasts a path query mes-
sage, which includes the IDs of the destination BS and the
destination node, to the mobile nodes in its region. The path
querying process is similar to the previous path querying
for a lightly loaded BS. The nodes further forward the path
query to their neighbors until the query reaches the destina-
tion node. Here, we do not piggyback the query to beacon
messages because this querying is for a specific mobile node
rather than any mobile node near a lightly loaded BS.
Including the mobile node’s ID into beacon messages gener-
ates very high overhead.
In order to reduce the broadcasting overhead, a mobile
node residing in the region of a BS not close to the destina-
tion BS drops the query. The nodes can determine their
approximate relative positions to BSes by sensing the signal
strengths from different BSes. Each node adds the strength
of its received signal into its beacon message that is periodi-
cally exchanged between neighbor nodes so that the nodes
can identify their relative positions to each other. Only those
mobile nodes that stay farther than the query forwarder
from the forwarder’s BS forward the queries in the direction
of the destination BS. In this way, the query can be for-
warded to the destination BS faster. After the multi-hop
path is discovered, the neighboring BS sends the segment to
the destination node along the path. Since the destination
node is in the neighboring BS’s region, the overhead to iden-
tify a path to the destination node is small. Note that our
methods for congestion control in base stations involve
query broadcasting. However, it is used only when some
base stations are overloaded rather than in the normal DTR
routing algorithm in order to avoid load congestion in BSes.
Fig. 4 shows an example of the congestion control on
BSes when b ¼ 2. As shown in figure, BS B1 is congested.
Then, the relay nodes of the source node’s message broad-
cast locally by beacon piggybacking to find multi-hop paths
which lead to B3 and B4. The relay nodes then send seg-
ments along the paths. In this way, the traffic originally tar-
geting overloaded B1 can be spread out to the neighboring
BSes B3 and B4. B3 and B4 further forward the segments to
the destination BS B6 if B6 is not congested. If B6 is also con-
gested, B3 and B4 send the segments to the neighboring
BSes of B6. Specifically, B4 sends the segment to B3. B3 does
not forward the segment to another BS since it already is
close to B6. B3 then finds a multi-hop path to the destination
node and uses ad-hoc transmission to forward the segments
to the destination node. Similarly, when B2 wants to send a
segment to the destination node, it also uses a multi-hop
path for the segment transmission.
4 PERFORMANCE ANALYSIS OF THE DTR
PROTOCOL
In this section, we analyze the effectiveness of the DTR pro-
tocol at enhancing the capacity and scalability of hybrid
wireless networks. In our analysis, we use the same scenario
in [17] for hybrid wireless networks, and use the same sce-
nario in [33] for the ad-hoc network component. We present
the scenarios and some concepts below. We consider a large
number of mobile nodes uniformly and randomly deployed
over a 2D field. The moving directions of the nodes are
independent and identically distributed (i.i.d.). The distri-
bution of mobile nodes can be modeled as a homogeneous
Poisson process with node density s [34]. That is, given an
area of size S in the field, the number of nodes in the area,
denoted by nðSÞ, follows the Poisson distribution with the
parameter sS,
Pr nðSÞ ¼ kð Þ ¼
ðsSÞk
eÀsS
k!
; k ¼ 0; 1; 2; . . . (1)
Fig. 4. Congestion control on BSes.
8143271457
8143271457

Besides mobile nodes, there are M BSes regularly deployed
in the field. The BSes divide the area into a hexagon tessella-
tion, in which each hexagon has side length h. The BSes are
assumed to be connected together by a wired network. We
assume that the link bandwidths in the wired network are
large enough so that there are no bandwidth constraints
between BSes. In single-path transmission, a message is
sequentially transmitted in one routing path. In multi-path
transmission, a message is divided into a number of seg-
ments that are forwarded along multiple paths in a distrib-
uted manner. We assume each segment has the same length
l. Table 1 lists the notations used in our analysis.
We assume that the transmission range of all mobile
nodes and all BSes is R (R h). In this paper, we use proto-
col model [17], [33] to describe the interference among nodes;
that is, a transmission from a node (here “node” can be
either mobile node or BS) vi to another node vj is successful
if the following two conditions are satisfied: 1) vj is within
the transmission range of vi, i.e.,
jvi À vjj R; (2)
where jvi À vjj represents the euclidean distance between vi
and vj in the plane. 2) For any other node vk that is simulta-
neously transmitting over the same channel,
jvk À vjj ! ð1 þ DÞjvi À vjj: (3)
Formula (3) guarantees a guard zone around the receiving
node to prevent a neighboring node from transmitting on
the same channel at the same time. The radius of the guard
zone is ð1 þ DÞ times the distance between the sender and
the receiver. The parameter D defines the size of the guard
zone and we require that D 0.
We first adopt a concept called aggregate throughput capac-
ity introduced in [17], [33] to measure the throughput of the
network.
Definition (Aggregate Throughput Capacity of Hybrid
Networks). The aggregate throughput capacity of a hybrid
wireless network is of order Qðfðs; MÞÞ if there are determin-
istic constants a 0, and a0
þ1 such that
lim
M!1
Pr Pðs; MÞ ¼ afðs; MÞ is feasibleð Þ ¼ 1 (4)
lim inf
M!1
Pr Pðs; MÞ ¼ a0
fðs; MÞ is feasibleð Þ 1: (5)
Since the working frequency of infrastructure networks is
around 700 MHz while that of ad-hoc networks is 2.4 GHz,
the communications in infrastructure mode (between
mobile nodes and BSes through cellular interface) would
not generate interference to ad-hoc mode. We divide the
channel for infrastructure mode transmissions into uplink
and downlink parts, according to the transmission direction
relative to the BSes. Accordingly, in the DTR protocol, the
traffic of each S-D pair is composed of at most two intra-cell
traffics, one uplink traffic and one download traffic. Since
uplink traffic and downlink traffic use different sub-chan-
nels, there is also no interference between these two types of
traffics. For each node vi, we denote the bandwidth assigned
to intra-cell, uplink, and downlink sub-channels by Wint
i ,
Wup
i and Wdown
i , respectively. We let Wup
i ¼ Wdown
i because
there are the same amount of uplink traffic and downlink
traffic. The transmission rates should sum to Wi, i.e.,
Wint
i þ Wup
i þ Wdown
i ¼ Wi. Though no interference exists
between intra-cell, uplink, and downlink traffics, interfer-
ence exists between the same type of traffic in a cell and
between different cells. Fortunately, there is an efficient spa-
tial transmission schedule that can prevent such interfer-
ences [17]. First, to avoid the interference in a cell, any two
nodes within the cell are not allowed to transmit with the
same traffic mode at the same time. Second, to avoid the
interference between different cells, the cells are spatially
divided into a number of groups and transmissions in the
cells of the same group do not interfere with each other. If
the groups are scheduled to transmit in a round robin fash-
ion, each cell will be able to transmit once every fixed
amount of time without interfering with each other.
Below, we show how many groups we need to divide the
cells to prevent interference. We adopt the notion of interfer-
ing neighbors introduced in [17], and give the number of cells
that can be affected by a transmission in one cell. Two cells
are defined to be interfering neighbors if there is a point in
one cell which is within a distance (2 þ D)R of a point in the
other cell. Accordingly, if two cells are not interfering neigh-
bors, transmissions in one cell do not interfere with trans-
missions in the other cell. [17] has proved that (1) each cell
has no more than c1 interfering neighbors (Lemma 1 in
[17]), where c1 is a constant
c1 ¼
4
3
3l þ 2R þ DR
3l
2
; (6)
and (2) all cells should be divided into c1 þ 1 groups and the
whole channel should be divided into c1 þ 1 subchannels,
where each subchannel is allocated to the cells in one group.
Thus, the number of group we need to divide the cells to
prevent interference is c1 þ 1.
Before calculating the aggregate throughput capacity of
DTR, we first introduce Lemma 4.1.
Lemma 4.1. The number of cells that have mobile nodes is QðMÞ.
Proof. Denote the number of cells having mobile nodes
by M1. To prove M1 ¼ QðMÞ, we need to prove that
there exists deterministic constants a 0 and a0
þ1
such that
lim
M!1
Pr M1 ¼ aMð Þ ¼ 1; (7)
lim inf
M!1
Pr M1 ¼ a0
Mð Þ 1: (8)
TABLE 1
Parameter Table
s Node density M Number of BSes
l Segment’s length sh Area size of a cell
nðSÞ Number of nodes
in area S
R Transmission range
Wi Bandwidth of a node vi mi Mobile node i
Pðs; MÞ Throughput nðs; MÞ Number of nodes
8143271457
8143271457

For Formula (8), let a0
¼ 2. Because the number of cells
having mobile nodes is upper bounded by M, then
lim inf
M!1
Pr M1 ¼ 2M is feasibleð Þ ¼ 0: (9)
Now, we prove that Formula (7) can also be satisfied
for some constant a. Because the number of nodes in a
cell follows a Poisson distribution and the size of each
cell (hexagon) is sh ¼ 3
ffiffiffi
3
p
h2
, then we can derive the
probability that no mobile node is in a cell equals
Pr nðShÞ ¼ 0ð Þ ¼
s0
eÀsh
0!
¼ eÀsh : (10)
Consider an arbitrary cell k, let X1; X2, . . ., Xk, . . ., XM be
i.i.d. random variables, where Xk represents whether cell
k has mobile nodes. Then, Xk is defined as follows:
Xk ¼
1 cell k has mobile nodes
0 cell k does not have mobile nodes

(11)
and EðXkÞ ¼ eÀsh . For simplicity, let c2 ¼ 1 À eÀsh . Then,
M1 ¼
PM
k¼1 Xk. By theStrong Law of Large Number
(SLLN) [34],
Pr lim
M!1
PM
k¼1 Xk
M
¼ c2
!
¼ 1; (12)
which implies that limM!1 Pr M1 ¼ c2Mð Þ ¼ 1, which
indicates that when a ¼ c2, Formula (7) can also be
satisfied. tu
Lemma 4.2. Let nðs; MÞ denote the number of mobile nodes in
the whole network. Then,
lim
M!1
Pr nðs; MÞ ¼ shMð Þ ¼ 1: (13)
Proof. Let Z1, Z2, . . . , ZM be i.i.d. random variables repre-
senting the number of nodes in cell 1, 2, . . . , M, respec-
tively. Then, nðs; MÞ ¼
PM
k¼1 Zk. Because each Zk follows
a Poisson distribution with parameter sh, EðZkÞ ¼ sh,
81 k M. According to SLLN,
Pr lim
M!1
PM
k¼1 Zk
M
¼ sh
!
¼ 1; (14)
which implies that limM!1 Prð
PM
k¼1 Zk ¼ shMÞ ¼ 1, and
hence limM!1 Pr nðs; MÞ ¼ shMð Þ ¼ 1. tu
Theorem 4.1. For a hybrid network of M BSes and s mobile
node density, where each node has the intra-cell, uplink and
downlink sub-channel bandwidth satisfying
Wdown
i ¼ Wup
i ¼ Wup
¼ W=4; Wint
i ¼ Wint
¼ W=2 (15)
the aggregate throughput capacity of DTR is
Pðs; MÞ ¼ QðMWÞ: (16)
Proof. To prove Pðs; MÞ ¼ QðMWÞ, we need to prove that
there exists deterministic constants a 0 and a0
1
such that
lim
M!1
PrfPðs; MÞ ¼ aMW is feasibleg ¼ 1 (17)
lim inf
M!1
PrfPðs; MÞ ¼ a0
MW is feasibleg 1: (18)
Recall that any two nodes within a cell cannot trans-
mit simultaneously in the same traffic mode, the
throughput P is upper bounded by MW=4, which can
be achieved only if each cell has one node to send the
message. Hence, Formula (18) can be satisfied by set-
ting a0
to 1/2.
Then, we will show how Formula (17) can be satis-
fied. Since the same message has to go through an
uplink and a downlink and it is counted only once in
the throughput capacity, calculating the throughput of
the whole network is equivalent to calculating the
throughput of uplink traffic Pup or the throughput of
downlink traffic Pdown. Notice calculating intra-cell
traffic throughput is not accurate because a message
may transmit twice with intra-cell mode. In this proof,
we calculate Pup.
First, we consider the throughput of the uplink traffic
of an arbitrary cell k, denoted by Pk
up. Since the schedule
allocates 1=ðc1 þ 1Þ time slots to this cell, then
Pk
up ¼
Wup
c1 þ 1
: (19)
Then, we consider the throughput of the whole network.
Let Pup ¼
PM
i¼1 Pi
upXi represent the throughput of uplink
traffic, then we have
lim
M!1
Pr Pup ¼
c2MW
3ðc1 þ 1Þ

¼ lim
M!1
Pr
XM
i¼1
Pi
upXi ¼
c2MWup
c1 þ 1
!
¼ lim
M!1
Pr
XM
i¼1
Xi ¼ c2M
!
¼ 1 ðBy Lemma 4:1Þ:
Accordingly, Formula (17) can be satisfied when a is
set to c2
3ðc1þ1Þ. tu
Corollary 4.1. With the restriction in Theorem 4.1, DTR can
achieve QðWÞ throughput per S-D pair.
Proof. Denote the throughput of per S-D pair by P, which
equals
P ¼
Pðs; MÞ
n
: (20)
Obviously, P is upper bounded by W
4 because each node
has at most W
4 for uplink traffic (or downlink traffic),
which equals its S-D pair throughput. By Lemma 4.2 and
Theorem 4.1, we can derive that
8143271457
8143271457

lim
M!1
Pr P ¼
c2W
3ðc1 þ 1Þsh

¼ lim
M!1
Pr
Pðs; MÞ
nðs; MÞ
¼
c2W
3ðc1 þ 1Þsh

! lim
M!1
Pr Pðs; MÞ ¼
c2WM
3ðc1 þ 1Þ

Pr nðs; MÞ ¼ shMð Þ
¼ lim
M!1
Pr Pðs; MÞ ¼
c2WM
3ðc1 þ 1Þ

¼ 1
which implies that limM!1 Pr P ¼ c2W
3ðc1þ1Þsh

¼ 1. tu
Corollary 4.1 shows that DTR produces a constant
throughput for each pair of nodes regardless of the num-
ber of nodes in each cell due to its spacial reuse of the
system. Theorem 4.1 and Corollary 4.1 show that the
aggregate throughput capacity and the throughput per
S-D pair of DTR are QðMWÞ and QðWÞ, respectively.
The work in [17] proves that DHybrid achieves QðMWÞ
infrastructure aggregate throughput, and the work in
[33] proves that the pure ad-hoc transmission achieves
Qð Wffiffiffiffiffiffiffiffiffiffi
nÁlogn
p Þ throughput per S-D pair. The results demon-
strate that the throughput rates of DTR and DHybrid are
higher than that of the pure ad-hoc transmission. This is
because the pure ad-hoc transmission is not efficient in a
large scale network [35]. A large network size reduces
the path utilization efficiency and increases node inter-
ference. Facilitated by the infrastructure network, DTR
and DHybrid avoid long distance transmissions, leading
to a higher transmission throughput.
Proposition 4.1. Suppose a mobile node needs to allocate totally
U segments with the same length to L neighboring mobile
nodes m1, . . ., mL, which has uplink bandwidth Wup
1 , . . . ,
Wup
L , respectively. Let Ui denote the number of segments to be
allocate to mi (i ¼ 1; 2; . . . ; L). To minimize the average
latency of these segments, the optimal allocation should satisfy
U1
W
up
1
¼ Á Á Á ¼ UL
W
up
L
. The minimized average latency equals
Ul
2
PL
i¼1
W
up
i
.
Proof. Recall that each segment has length l. Then, for
each mobile node mi it requires l
W
up
i
time to transmit a
segment. Therefore, the jth segment that mi needs to
transmit has to wait ðjÀ1Þl
W
up
1
slots. Hence, the total
latency of the segments that mi needs to transmit to
its BS equals
XUi
j¼1
ðj À 1Þl
Wup
i
¼
ð0 þ 1 þ Á Á Á þ ðUi À 1ÞÞl
Wup
i
%
U2
i l
2Wup
i
: (21)
Hence, the average latency of transmitting all the mes-
sages should be
PK
i¼1
U2
i
l
2W
up
i
=U. According to Cauchy-
Schwarz inequality [34], the average latency is lower
bounded
1
U
XL
i¼1
U2
i l
2Wup
i
¼
l
2U
PL
i¼1 Wup
i
XL
i¼1
U2
i
Wup
i
XL
i¼1
Wup
i
!
l
2U
PL
i¼1 Wup
i
XL
i¼1
ffiffiffiffiffiffiffiffiffi
U2
i
Wup
i
s
ffiffiffiffiffiffiffiffiffi
Wup
i
q
!2
¼
Ul
2
PL
i¼1 Wup
i
: (22)
When
ffiffiffiffiffiffiffi
U2
1
Wup
1
r
W
up
1
p ¼ Á Á Á ¼
U2
L
Wup
L
r
W
up
L
p , or equivalently, U1
W
up
1
¼ Á Á Á
¼ UL
W
up
L
, the average segment latency
PL
i¼1
U2
i l
2W
up
i
=U can
achieve the minimum value Ul
2
PL
i¼1
W
up
i
. tu
Proposition 4.1 indicates that forwarding segments to the
nearby nodes with the highest capacity can minimize the
average latency of messages in the cell. It also balances
the transmission load of the mobile nodes within a cell.
Proposition 4.2 A source node in DTR can find relay nodes for
message forwarding with probability
P1
k¼1
kÀ1
k
ck
r eÀcr
k! , where
cr ¼ pR2
.
Proof. Let m denote the number of nodes within mi’s trans-
mission area and define the indicator variable Qi by
Qi ¼
1 mi is the highest capacity node
0 mi is not the highest capacity node

(23)
then,
Prfmi can find relays for message forwardingg
¼
X1
k¼0
Pr Qi ¼ 0jm ¼ kð ÞPr m ¼ kð Þ ¼
X1
k¼1
k À 1
k
ck
reÀcr
k!
:
tu
Proposition 4.2 indicates that in a high-density network,
a source node in DTR can find relay nodes for message
forwarding with a high probability. For example, assume
the average number of neighbor nodes of a source node is
10. With the daily increasing number of mobile devices,
such an assumption is realistic. Then, the probability of
not being able to find any node in the range of a node is
1 À
P1
k¼1
kÀ1
k
10keÀ10
k! % 0:12, which is very small. Therefore,
in a high-density network, a source node can find neigh-
bors for message forwarding with a high probability.
We use DHybrid to denote the group of routing protocols
in hybrid wireless networks that directly combine the ad-
hoc transmission mode and the infrastructure transmission
mode [1], [5], [6], [12], [13], [14], [15], [16], [17], [18].
Proposition 4.3. In a hybrid wireless network, the DHybrid rout-
ing protocol leads to load imbalance among the mobile nodes in
a cell.
Proof. Fig. 5a shows a cell with a BS and a randomly picked
mobile node mi in the range of the BS. The shaded region
represents all possible positions of the source nodes that
8143271457
8143271457

choose mi as the relay node in DHybrid. The total traffic
passing through node mi is the sum of the traffic gener-
ated by the nodes in the shaded region. The area of
shaded region is
S ¼ sh À pD2
ð0 D hÞ; (24)
where D is the distance between the BS and relay node
mi and sh is the area size of a cell. Therefore, the expected
value of traffic passing through node mi is
W Á s Á ðsh À pD2
Þ ð0 D hÞ; (25)
where W is the data transmission rate of a source
node, and s is the density of the nodes in a region.
Equation (25) shows that the traffic passing through
node mi decreases as D increases. That is, the nodes
closer to the BS have a higher load than the nodes stay-
ing at the brim of the cell. tu
Proposition 4.4. In a hybrid wireless network, DTR achieves
more balanced load distribution among the mobile nodes in
each cell.
Proof. The shaded region in Fig. 5b represents all possible
positions of the source and relay nodes that choose node
mi as relay node. Suppose m neighbor nodes are chosen
as relay nodes, then the expected traffic passing through
node mi is W
m Á s Á pR2
which shows that the traffic going
through node mi is independent of its location relative to
its BS. Since every node in the cell has an equal probabil-
ity of generating traffic, the traffic load is balanced
among the nodes in the cell. tu
5 PERFORMANCE EVALUATION
This section demonstrates the properties of DTR through
simulations on NS-2 [36] in comparison to DHybrid [17],
Two-hop [19] and AODV [8]. In DHybrid, a node first uses
broadcasting to observe a multi-hop path to its own BS
and then forwards a message in the ad-hoc transmission
mode along the path. During the routing process, if the
transmission rate (i.e., bandwidth) of the next hop to the
BS is lower than a threshold, rather than forwarding
the message to the neighbor, the node forwards the mes-
sage directly to its BS. The source node will be notified if
an established path is broken during data transmission. If
a source sends a message to the same destination next
time, it uses the previously established path if it is not bro-
ken. In the Two-hop protocol, a source node selects the
better transmission mode between direct transmission and
relay transmission. If the source node can find a neighbor
that has higher bandwidth to the BS than itself, it transmits
the message to the neighbor. Otherwise, it directly trans-
mits the message to the BS.
Unless otherwise specified, the simulated network con-
sists of 50 mobile nodes and four BSes. In the ad-hoc compo-
nent of the hybrid wireless network, mobile nodes are
randomly deployed around the BSes in a field of
1;000 Â 1;000 square meters. We used the Distributed Coor-
dination Function (DCF) of the IEEE 802:11 as the MAC layer
protocol. The transmission range of the cellular interface
was set to 250 meters, and the raw physical link bandwidth
was set to 2 Mbits/s. The transmission power of the ad-hoc
interface was set to the minimum value required to keep the
network connected for most times, even when nodes are in
motion in the network. Then, the influence of the transmis-
sion range on different methods’ performance is controlled.
Specifically, we set the transmission range through the ad-
hoc interface to 1.5 times of the average distance between
neighboring nodes, which can be obtained by measuring the
simulated network. We used the two-ray propagation model
for the physical layer model. Constant bit rate (CBR) was
selected as the traffic mode in the experiment with a rate of
640 kbps. In the experiment, we randomly chose four source
nodes to continuously send messages to randomly chosen
destination nodes. The number of channels for each BS was
set to 10. We set the number of redundant routing paths b in
Sections 3.4 to 1. We assumed that there was no capacity
degradation during transmission between BSes. This
assumption is realistic considering the advanced technolo-
gies and hardware presently used in wired infrastructure
networks. There was no message retransmission for failed
transmissions in the experiments.
We employed the random way-point mobility model [37]
to generate the moving direction, speed, and pause duration
of each node. In this model, each node moves to a random
position with a speed randomly chosen from ð1 À 20Þ m/s.
The pause time of each node was set to 0. We set the number
of segments of a message to the connection degree of the
source node. The simulation warmup time was set to 100 s
and the simulation time was set to 1,000 s. We conducted
the experiments five times and used the average value as
the final experimental result. To make the methods compa-
rable, we did not use the congestion control algorithm in
DTR unless otherwise indicated.
5.1 Scalability
Fig. 6 shows the average throughput measured in kbps per
S-D pair of different routing protocols versus the number of
mobile nodes in the system. The figure shows the through-
put of DTR remains almost the same with different network
sizes. This result conforms to Corollary 4.1. DTR uses dis-
tributed multi-path routing to fully take advantage of the
spatial reuse and avoid transmission congestion in a single
path. Unlike the multi-hop routing in mobile ad-hoc net-
works, DTR does not need path query and maintenance.
Also, it limits the path length to three to avoid problems in
long-path transmission. The throughput of DHybrid and
AODV decreases as the number of nodes in the network
increases. This is mainly because when the network size
increases, more beacon messages are generated in the
Fig. 5. The traffic load in DHybrid and DTR.
8143271457
8143271457

network. Also, the long transmission path also leads to high
transmission interference. Then, nodes in these methods
suffer from intense interference, leading to more transmis-
sion failure and degraded overall throughput. Also, the
mobile node increase in the system leads to high network
dynamism, resulting in frequent route re-establishments.
The short routing paths in Two-hop reduce congestion
and signal interference, thus enabling better spatial reuse as
in DTR. Meanwhile, Two-hop enables nodes to adaptively
switch between direct transmission and relay transmission.
Hence, part of the transmission load is transferred to relay
nodes, which carry the messages until meeting the BSes. As
a result, gateway nodes connecting mobile nodes and BSes
are not easily overloaded. Therefore, the throughput of
Two-hop is higher than DHybrid. However, since the num-
ber of message routing hops is confined to one, Two-hop
may not find the node with the best transmission rate to the
BSes because of the short transmission range of the ad-hoc
interface. Therefore, the throughput of Two-hop is lower
than DTR, especially in a network with high node density.
The reason that AODV has the lowest throughput per S-D
pair is its long transmission paths.
Fig. 7 shows the throughput per S-D pair versus the
number of BSes in different routing protocols. The number
of BSes was varied from 3 to 6. The BSes are uniformly dis-
tributed in the network. We can see from the figure that as
the number of BSes increases, the throughputs of DTR,
Two-hop, and DHybrid increase while the throughput of
AODV stays nearly constant. In DTR, Two-hop, and DHy-
brid, as the number of BSes increases, the total number of
nodes close to the BSes increases. Then, more nodes have
high transmission rates to the BSes, leading to a throughput
increase. In AODV, since the traffic between S-D pairs does
not travel though BSes, the throughput between an S-D pair
is not affected by the increased number of BSes in the net-
work. The figure also shows that the throughput of DTR is
constantly larger than Two-hop and the throughput of
Two-hop is constantly larger than DHybrid. AODV con-
stantly has the lowest transmission delay due to the same
reasons as in Fig. 6.
5.2 Transmission Delay
Fig. 8 shows the average transmission delay of S-D pairs
for successfully delivered messages in different routing
protocols versus network size. The network size was var-
ied from 20 to 100 with 20 increase in each step. Transmis-
sion delay is the amount of time it takes for a message to
be transmitted from its source node to its destination
node. From the figure, we see that DTR generates the
smallest delay. In DTR, each source node first divides
its messages into smaller segments and then forwards
them to the nearby nodes with the highest capacity, which
leads to more balanced transmission load distribution
among nodes than the previous methods. According to
Proposition 4.1, average latency can be minimized when
the transmission loads of all the nodes are balanced.
Hence, DTR has smaller latency than the previous meth-
ods. The delay of DHybrid is five to six times larger than
DTR. DHybrid uses a single transmission path, while DTR
uses multiple paths. Recall that we set the number of seg-
ments of a message to the connection degree of the source
node in DTR. Thus, the ratio of delay time of DHybrid to
that of DTR equals the average connection degree. As the
number of nodes in the system increases, the connection
degree of each node increases, and the increase rate of the
ratio grows. This is caused by two reasons. First, a higher
node density leads to longer path lengths in DHybrid,
resulting in a longer delay because of a higher likelihood
of link breaks. Second, a higher node density enables a
node to quickly find relay nodes to forward messages in
DTR, as indicated in Proposition 4.2.
DTR also produces a shorter transmission delay than
Two-hop for two reasons. First, the multi-path parallel rout-
ing of DTR saves much transmission time as shown in
Proposition 4.1. Second, the distributed routing of DTR ena-
bles some messages to be forwarded to the destination BS’s
neighboring cells with high transmission rates rather than
waiting in the current hot cell for a transmission channel.
We can also observe that Two-hop produces lower delay
than DHybrid. This is because the delay of DHybird
includes the time for establishing a path and for data trans-
mission. Also, the multi-hop transmission component of
DHybrid results in a higher delay due to the queuing delay
in each hop. Because of the long distance transmissions
without support from an infrastructure network, AODV
generates the longest delay.
Fig. 6. Throughput vs. network size (simulation).
Fig. 7. Throughput vs. number of BSes.
Fig. 8. Delay vs. network size.
8143271457
8143271457

Fig. 9 plots the average communication delay per S-D
pair for successfully delivered messages versus the number
of BSes in different routing protocols. The figure shows that
the increasing number of BSes in the system leads to a com-
munication delay decrease between nodes in DTR, Two-
hop, and DHybrid, but does not affect the communication
delay in AODV. In DTR, Two-hop, and DHybrid, as the
number of BSes increases, more nodes can stay close to the
BSes, leading to fewer communication hops and better
transmission links between nodes and BSes. Thus, the trans-
mission delay between the nodes is reduced. Since the
communication between S-D pairs in AODV does not
rely on BSes, AODV maintains constant communication
delay. The figure also shows that the communication delay
between S-D pairs follows DTRTwo-hopDHy-
bridAODV for the same reason as in Fig. 8.
5.3 Communication Overhead
We use the generation rate of control messages in the net-
work and MAC layers in kbps to represent the communica-
tion overhead of the routing protocols. Fig. 10 illustrates the
communication overhead of DTR, Two-hop, DHybrid, and
AODV versus network size. We can see that the communi-
cation overheads of DTR and Two-hop are very close. This
is because both DTR and Two-hop are transmission proto-
cols of short distance and small hops. DTR has slightly
higher communication overhead than Two-hop because
DTR utilizes three hop transmission, which has one more
hop than two hop transmission. However, the marginal
overhead increase leads to a much higher transmission
throughput as shown in Fig. 6. DHybrid generates much
higher overhead than DTR and Two-hop because of the
high overhead of routing path querying. The pure AODV
routing protocol results in much more overhead than the
others. This is because without an infrastructure network,
the messages in AODV travel a long way from the source
node to the destination node through much longer paths.
5.4 Effect of Mobility
In order to see how the node mobility influences the perfor-
mance of the routing protocols, we evaluated the through-
put of these four transmission protocols with different node
mobilities. Fig. 11 plots the throughput of DTR, DHybrid,
Two-hop, and AODV versus node moving speed. From the
figure, we can see that the increasing mobility of the nodes
does not adversely affect the performance of DTR and Two-
hop. It is intriguing to find that high mobility can even help
DTR to increase its throughput and that Two-hop generates
constant throughput regardless of the mobility. This is
because the DTR and Two-hop transmission modes do not
need to query and rely on multi-hop paths; thus, they are
not affected by the network partition and topology changes.
Moreover, since DTR transmits segments of a message in a
distributed manner, as the mobility increases, a mobile
node can meet more nodes in a shorter time period. There-
fore, DTR enables the segments to be quickly sent to high-
capacity nodes. As node mobility increases, the throughput
of DHybrid decreases. In DHybrid, the messages are routed
in a multi-hop fashion. When the links between nodes are
broken because of node mobility, the messages are dropped.
Therefore, when nodes have smaller mobility, the links
between the mobile nodes last longer and more messages
can be transmitted. Hence, the throughput of DHybrid is
adversely affected by node mobility. However, since DHy-
brid can adaptively adjust the routing between the ad-hoc
transmission and cellular transmission, the throughput of
DHybrid is much higher than AODV’s. With no infrastruc-
ture network, AODV produces much lower throughput
than the others. Its throughput also drops as node mobility
increases for the same reasons as DHybrid.
5.5 Effect of Workload
We measured the total throughput of BSes on the messages
received by BSes. Fig. 12 shows the total throughput of the
BSes versus the number of source nodes. We can see that
DTR and Two-hop have much higher throughput increase
rates than DHybrid. This is because in DTR and Two-hop,
the number of transmission hops from a source node to a BS
is small. Meanwhile, each node can adaptively switch
between relay transmission and direct transmission based
on the transmission rate of its neighbors. Hence, part of a
source node’s transmission load is transferred to a few relay
nodes, which carry the messages until meeting the BSes.
Therefore, the gateway mobile nodes are less likely to be
congested. However, nodes in DHybrid cannot adaptively
adjust the next forwarding hop because it is predetermined
in the routing path. Messages are always forwarded to the
Fig. 9. Delay vs. number of BSes.
Fig. 10. Overhead vs. network size.
Fig. 11. Throughput vs. mobility.
8143271457
8143271457

mobile gateway nodes that are closer to the BSes or that
have higher transmission rates. Therefore, these mobile
gateway nodes can easily become congested as the work-
load of the system increases, leading to many message
drops. Therefore, when the number of the source nodes is
larger than 4, the throughput of DHybrid remains nearly
constant. This is also the reason that the throughput of DHy-
brid is constantly lower than those of DTR and Two-hop.
Additionally, the figure shows that the overall throughput
of Two-hop is lower than that of DTR. This is because most
of the traffic in Two-hop is confined to a single cell. When a
BS in a cell is congested, the traffic cannot be transferred to
other cells. In contrast, DTR’s three-hop distributed for-
warding mechanism enables it to distribute the traffic
among the BSes in a balance. Therefore, the BSes in DTR
will not become congested easily. In addition, as the for-
warding mechanism gives nodes more flexibility in choos-
ing relay nodes with higher transmission rates for message
forwarding to the BSes, the overall BS throughput in DTR is
larger than in Two-hop.
5.6 Effect of the Number of Routing Hops
We conducted experiments to show the optimal number of
routing hops for the routing in hybrid wireless networks.
We tested the throughput per S-D pair for x-hop DTR,
where x was varied from 1 to 4. In the one-hop routing, a
node directly transmits a message to the BS without mes-
sage division. In the other routing protocols, the ðx À 1Þth
hop chooses the best transmission mode between direct
transmission and relay transmission. Also, in the four-hop
routing, the second relay node randomly chooses the third
relay node.
Fig. 13 shows the average throughput per S-D pair versus
network size in DTR. As the figure shows, as the network
size increases, the node throughput keeps constant regard-
less of the number of forwarding hops in a routing. The rea-
son is the same as in Fig. 6. We can also see from the figure
that the throughput of the four protocols follows 3-hop4-
hop2-hop1-hop. In the one-hop routing, each node only
transmits segments directly to a BS regardless of its current
transmission rate. In the two-hop routing, if the transmis-
sion rate of a node’s neighbor is higher than that of the
node, it asks its neighbor node to forward the segment to a
BS. Therefore, the two-hop routing has higher throughput
than the one-hop routing. The three-hop routing can greatly
increase the number of node options for segment routing
since the number of nodes that the source node can encoun-
ter increases from d to d2
, where d is the average node
degree. Thus, a node with a greater transmission rate can be
chosen as the forwarding node. Meanwhile, the three-hop
routing can greatly facilitate inter-cell communication
because a node has a higher probability of reaching a
neighboring BS within a three-hop path length than within
a two-hop path length. Therefore, the throughput of the
three-hop routing is much higher than that of the two-hop
routing. The figure also shows that the four-hop routing
produces lower throughput than the three-hop routing. The
reason is that three hops are enough to find a hop with high
transmission rate and achieve inter-cell communication
because of widespread BSes. The four-hop routing increases
the forwarding delay due to the greater number of nodes in
a route; thus, it cannot increase the uploading transmission
rate of messages.
5.7 Load Distribution within a Cell
In this experiment, we tested the load distribution of mobile
nodes in a randomly chosen cell in the hybrid wireless net-
work that employs each of the DTR, DHybrid, and Two-
hop protocols. We normalized the distance from a mobile
node to its base station according to the function D
Rb
, where
D is the actual distance and Rb is the radius of its cell. We
divided the space of the cell into several concentric circles
and measured the loads of the nodes on each circle to show
the load distribution.
Fig. 14 shows the average load of a node corresponding
to the normalized distance from itself to the BS in the chosen
cell. The figure shows that most of the traffic load of DHy-
brid is located at nodes near the BS. The nodes far from the
BS have very low load. The results conform to Proposition
4.3. In DHybrid, if a source node wants to access the Internet
backbone or engage in inter-cell communication, it trans-
mits the messages to the BSes in a multi-hop fashion. There-
fore, the nodes near the BSes will have the highest load. On
the other hand, since there is little traffic going through the
nodes at the brim of a cell, the load of these nodes is small.
As a result, some nodes can easily become hot spots while
Fig. 12. Throughput of BSes vs. number of source nodes.
Fig. 13. Throughput vs. number of hops.
Fig. 14. Load distribution in a cell.
8143271457
8143271457

the resources of other nodes are not fully utilized. This load
imbalance prevents DHybrid from fully utilizing system
resources. The traffic load of DTR is almost evenly distrib-
uted in the system, which is in line with Proposition 4.4. In
DTR, the traffic from a source node is distributed among a
number of relay neighbors for further data forwarding. The
nodes at the brim of the cell also take responsibility for mes-
sage forwarding, since the neighbor nodes of the brim
nodes could be located in other cells with good transmission
channels. In Two-hop, the source node considers direct
transmission or one-hop relay transmission based on the
channel condition. Since the node is chosen within one hop,
the messages will not gather close to the BS due to the lim-
ited transmission range. However, because of its sequential
transmission, Two-hop cannot achieve load balance among
nodes in a cell as well as DTR.
5.8 Load Balance between Cells
In this experiment, we tested the effectiveness of the conges-
tion control algorithm in DTR. We also added a congestion
control algorithm to DHybrid. In the algorithm, when a
node receives beacon messages from its BS indicating that it
is overloaded, the node broadcasts a query message to find
a path to a nearby uncongested BS. We selected two BSes
out of the total four BSes. In the range of each of the two
selected BSes, we randomly selected one mobile node as the
source node to send messages to a randomly selected desti-
nation node in the network. Once the source node moves
out of the range of the selected BS, another mobile node in
the range was selected as the source node. In order to show
the load distribution of the BSes in different protocols, we
ranked the BSes based on BS throughput. The BS with the
highest throughput has a rank of 1.
Fig. 15 shows the throughput of each BS versus the BS
rank. We can see from the figure that in Two-hop, the
throughput of the first two BSes is extremely high while
the throughput of the last two BSes is extremely small.
This is because the two hop routing path length in Two-
hop is not long enough to forward messages from a con-
gested BS to a lightly loaded BS. Therefore, the traffic can-
not be shifted to the neighboring lightly loaded BSes,
leading to an unbalanced load distribution. We can also
see from the figure that in DTR, the variance of the
throughputs in different BSes is small. The reason is that
three forwarding hops are enough for a mobile node to
reach a neighboring BS and hence to balance the load
between the BSes. Meanwhile, the congestion control algo-
rithm in DTR can effectively switch the traffic from a
highly loaded cell to a lightly loaded cell. Because the
BSes of ranks 1 and 2 in DTR are not congested, their
throughput is less than the corresponding BSes in Two-
hop; also, the throughput of the BSes of ranks 3 and 4 in
DTR is much higher than that of the corresponding BSes
in Two-hop. DHybrid achieves more balanced load distri-
bution between BSes than Two-hop since it employs a con-
gestion control algorithm. In DHybrid, if a previously
established path to a destination is not broken, a node still
uses this path to transmit messages to the same destina-
tion. Thus, the nodes cannot dynamically balance load
between BSes. Also, when a node finds that its current BS
is congested, it takes a long time for it to find a path to a
non-congested BS by re-issuing a query message to the
neighboring non-congested BS, which greatly reduces the
throughput of the system.
Fig. 16 further shows the throughput of the BSes versus
simulation time in the three routing protocols. At the begin-
ning, the BSes with ranks 1 and 2 are congested and those
with ranks 3 and 4 do not have much traffic. Thus, the three
figures show that the BSes with ranks 1 and 2 have high
throughput but those with ranks 3 and 4 have extremely
low throughput at the beginning in all three protocols.
Fig. 16a shows the throughput of the BSes in DTR. As
shown in the figure, since DTR can adaptively adjust the
traffic among the BSes using its congestion control algo-
rithm, the throughput of the two highly congested BSes is
distributed to the neighboring BSes. As the traffic is for-
warded from the BSes of ranks 1 and 2 to the BSes of ranks 3
and 4, the throughputs of these BSes are very similar later in
the simulation. This result indicates the effectiveness of the
congestion control algorithm in DTR for load balance
between cells.
Fig. 16b shows the throughput of the BSes in Two-hop. In
Two-hop, since the source nodes cannot effectively move
the traffic between BSes, the BSes with rank 1 and rank 2
Fig. 15. Load distribution among BSes.
Fig. 16. Base station load vs. simulation time.
8143271457
8143271457

constantly have the highest throughput, while the BSes with
rank 3 and rank 4 constantly have low throughput. The low
throughput is produced when the immediate neighbors of
the source node are in the range of the neighboring BSes of
the source node’s BS. However, the probability of such cases
is very small. Fig. 16c shows the throughput of the BSes in
DHybrid. As the nodes in DHybrid cannot effectively bal-
ance the load between the BSes, the throughput of the BSes
of rank 1 and rank 2 is much larger than that of the BSes of
rank 3 and rank 4. Comparing Figs. 16b and 16c, we can
find that the throughput in DHybrid is lower than that in
Two-hop. This is because the multi-hop transmission in the
ad-hoc network in DHybrid greatly reduces the throughput.
Meanwhile, the mobile gateway nodes in DHybrid easily
become congested, leading to more message drops.
6 CONCLUSIONS
Hybrid wireless networks have been receiving increasing
attention in recent years. A hybrid wireless network com-
bining an infrastructure wireless network and a mobile ad-
hoc network leverages their advantages to increase the
throughput capacity of the system. However, current
hybrid wireless networks simply combine the routing pro-
tocols in the two types of networks for data transmission,
which prevents them from achieving higher system capac-
ity. In this paper, we propose a Distributed Three-hop Rout-
ing data routing protocol that integrates the dual features of
hybrid wireless networks in the data transmission process.
In DTR, a source node divides a message stream into seg-
ments and transmits them to its mobile neighbors, which
further forward the segments to their destination through
an infrastructure network. DTR limits the routing path
length to three, and always arranges for high-capacity
nodes to forward data. Unlike most existing routing proto-
cols, DTR produces significantly lower overhead by elimi-
nating route discovery and maintenance. In addition,
its distinguishing characteristics of short path length, short-
distance transmission, and balanced load distribution pro-
vide high routing reliability and efficiency. DTR also has a
congestion control algorithm to avoid load congestion in
BSes in the case of unbalanced traffic distributions in net-
works. Theoretical analysis and simulation results show
that DTR can dramatically improve the throughput capacity
and scalability of hybrid wireless networks due to its high
scalability, efficiency, and reliability and low overhead.
ACKNOWLEDGMENTS
This research was supported in part by US NSF grants NSF-
1404981, IIS-1354123, CNS-1254006, CNS-1249603, and
CNS-1025652, and Microsoft Research Faculty Fellowship
8300751. The authors would like to thank Mr. Kang Chen
for his help in addressing review comments. An early ver-
sion of this work was presented in the Proceedings of ICPP
2009 [38]. Haiying Shen is the corresponding author.
REFERENCES
[1] H. Luo, R. Ramjee, P. Sinha, L. Li, and S. Lu, “Ucan: A unified cell
and ad-hoc network architecture,” in Proc. 9th Annu. Int. Conf.
Mobile Comput. Netw., 2003, pp. 353–367.
[2] P. K. McKinley, H. Xu, A. H. Esfahanian, and L. M. Ni, “Unicast-
based multicast communication in wormhole-routed direct
networks,” IEEE Trans. Parallel Distrib., vol. 5, no. 12, pp. 1252–
1265, Dec. 1994.
[3] H. Wu, C. Qiao, S. De, and O. Tonguz, “Integrated cell and ad hoc
relaying systems: iCAR,” IEEE J. Sel. Areas Commun., vol. 19,
no. 10, pp. 2105–2115, Oct. 2001.
[4] Y. H. Tam, H. S. Hassanein, S. G. Akl, and R. Benkoczi, “Optimal
multi-hop cellular architecture for wireless communications,” in
Proc. Local Comput. Netw., 2006, pp. 738–745.
[5] Y. D. Lin and Y. C. Hsu, “Multi-hop cellular: A new architecture
for wireless communications,” in Proc. IEEE Conf. Comput. Com-
mun., 2000, pp. 1273–1282.
[6] P. Thiran, O. Dousse, and M. Hasler, “Connectivity in ad hoc and
hybrid networks,” in Proc. IEEE Conf. Comput. Commun., 2002,
pp. 1079–1088.
[7] E. P. Charles and P. Bhagwat, “Highly dynamic destination
sequenced distance vector routing (DSDV) for mobile computers,”
in Proc. Conf. Appl., Technol., Archit., Protocols Comput. Commun.,
1994, pp. 234–244.
[8] C. Perkins, E. Belding-Royer, and S. Das, “Ad hoc on demand dis-
tance vector (AODV) routing,” Internet Eng. Task Force, Univ.
California, Santa Barbara, CA, Tech. Rep. RFC 3561, USA, 2003.
[9] D. Johnson and D. Maltz, “Dynamic source routing in ad hoc
wireless networks,” T. Imielinski and H. Korth, eds. Mobile Com-
puting, volume 353 of The Kluwer International Series in Engi-
neering and Computer Science, Springer US, pp. 153–181, 1996.
[10] V. D. Park and M. Scott Corson, “A highly adaptive distributed
routing algorithm for mobile wireless networks,” in Proc. IEEE
Conf. Comput. Commun., 1997, p. 1405.
[11] R. S. Chang, W. Y. Chen, and Y. F. Wen, “Hybrid wireless network
protocols,” IEEE Trans. Veh. Technol., vol. 52, no. 4, pp. 1099–1109,
Jul. 2003.
[12] G. N. Aggelou and R. Tafazolli, “On the relaying capacity of next-
generation GSM cellular networks,” IEEE Pers. Commun. Mag.,
vol. 8, no. 1, pp. 40–47, Feb. 2001.
[13] T. Rouse, I. Band, and S. McLaughlin, “Capacity and power inves-
tigation of opportunity driven multiple access (ODMA) networks
in TDD-CDMA based systems,” in Proc. IEEE Int. Conf. Commun.,
2002, pp. 3202–3206.
[14] H. Y. Hsieh and R. Sivakumar, “On using the ad-hoc network
model in wireless packet data networks,” in Proc. 3rd Int. Symp.
Mobile Ad Hoc Netw. Comput., 2002, pp. 36–47.
[15] L. M. Feeney, B. Cetin, D. Hollos, M. Kubisch, S. Mengesha, and
H. Karl, “Multi-rate relaying for performance improvement in
IEEE 802.11 WLANs,” in Proc. 5th Int. Conf. Wired/Wireless Internet
Commun., 2007, pp. 201–212.
[16] J. Cho and Z. J. Haas, “On the throughput enhancement of the
downstream channel in cellular radio networks through multihop
relaying,” IEEE J. Sel. Areas Commun., vol. 22, no. 7, pp. 1206–1219,
Sep. 2006.
[17] B. Liu, Z. Liu, and D. Towsley, “On the capacity of hybrid wireless
networks,” in Proc. IEEE Conf. Comput. Commun., 2003, pp. 1543–
1552.
[18] H. Y. Hsieh and R. Sivakumar, “A hybrid network model for wire-
less packet data networks,” in Proc. IEEE Global Telecommun. Conf.,
2002, pp. 961–966.
[19] Y. Wei and D. Gitlin, “Two-hop-relay architecture for next-gener-
ation WWAN/WLAN integration,” IEEE Wireless Commun.,
vol. 11, no. 2, pp. 24–30, Apr. 2004.
[20] X. J. Li, B. C. Seet, and P. H. J. Chong, “Multihop cellular net-
works: Technology and economics,” Comput. Netw., vol. 52, no. 9,
pp. 1825–1837, 2008.
[21] B. Bengfort, W. Zhang, and X. Du, “Efficient resource allocation in
hybrid wireless networks,” in Proc. Wireless Commun. Netw. Conf.,
2011, pp. 820–825.
[22] P. Thulasiraman and X. Shen, “Interference aware resource alloca-
tion for hybrid hierarchical wireless networks,” Comput. Netw.,
vol. 54, no. 13, pp. 2271–2280, 2010.
[23] K. Akkarajitsakul, E. Hossain, and D. Niyato, “Cooperative packet
delivery in hybrid wireless mobile networks: A coalitional game
approach,” IEEE Trans. Mobile Comput., vol. 12, no. 5, pp. 840–854,
May 2013.
[24] T. Liu, M. Rong, P. Li, D. Yu, Y. Xue, and E. Schulz, “Radio
resource allocation in two-hop cellular relaying network,” in Proc.
Veh. Technol. Conf., 2006, pp. 91–95.
8143271457
8143271457

[25] T. Liu, M. Rong, H. Shi, D. Yu, Y. Xue, and E. Schulz, “Reuse parti-
tioning in fixed two-hop cellular relaying network,” in Proc. Wire-
less Commun. Netw. Conf., 2006, pp. 177–182.
[26] L. Guan, J. Zhang, J. Li, G. Liu, and P. Zhang, “Spectral efficient
frequency allocation scheme in multihop cellular network,” in
Proc. Veh. Technol. Conf., 2007, pp. 1446–1450.
[27] D. M. Shila, Y. Cheng, and T. Anjali, “Throughput and delay anal-
ysis of hybrid wireless networks with multi-hop uplinks,” in Proc.
IEEE Conf. Comput. Commun., 2011, pp. 1476–1484.
[28] B. Liu, P. Thiran, and D. Towsley, “Capacity of a wireless ad hoc
network with infrastructure,” in Proc. 8th ACM Int. Symp. Mobile
Ad Hoc Netw. Comput., 2007, pp. 239–246.
[29] C. Wang, X. Li, C. Jiang, S. Tang, and Y. Liu, “Multicast through-
put for hybrid wireless networks under gaussian channel model,”
IEEE Trans. Mobile Comput., vol. 10, no. 6, pp. 839–852, Jun. 2011.
[30] A. G. Valko, “Cellular ip: A new approach to internet host mobi-
lity,” ACM Comput. Commun., vol. 29, no. 1, pp. 50–65, 1999.
[31] C. Sarr, C. Chaudet, G. Chelius, and I. G. Lassous, “A node-based
available bandwidth evaluation in IEEE 802.11 ad hoc networks,”
Int. J. Parallel Emergent Distrib. Syst., vol. 21, no. 6, pp. 423–440,
2006.
[32] X. P. Costa, M. T. Moreno, and H. Hartenstein, “A simulation
study on the performance of hierarchical mobile ipv6,” in Proc.
18th Int. Teletraffic Congr., 2003, pp. 468–474.
[33] P. Gupta and P. R. Kumar, “The capacity of wireless networks,”
IEEE Trans. Inf. Theory, vol. 46, no. 2, pp. 388–404, Mar. 2000.
[34] L. B. Koralov and Y. G. Sinai, Theory of Probability and Random Pro-
cesses. New York, NY, USA: Springer, 2007.
[35] H. Y. Hsieh and R. Sivakumar, “Performance comparison of cellu-
lar and multi-hop wireless networks: A quantitative study,” in
Proc. SIGMETRIC Int. Conf. Meas. Model. Comput. Syst., 2001,
pp. 113–122.
[36] The network simulator-ns-2 [Online]. Available: http://www.isi.
edu/nsnam/ns/, 2011.
[37] M. Grossglauser and D. Tse, “Mobility increases the capacity of ad
hoc wireless networks,” IEEE Trans. Netw., vol. 10, no. 4, pp. 477–
486, Aug. 2002.
[38] Z. Li and H. Shen, “A distributed three-hop routing protocol to
increase the capacity of hybrid networks,” in Proc. Int. Conf. Paral-
lel Process., 2009, pp. 277–284.
Haiying Shen received the BS degree in com-
puter science and engineering from Tongji Uni-
versity, China, in 2000, and the MS and PhD
degrees in computer engineering from Wayne
State University in 2004 and 2006, respectively.
She is currently an associate professor in the
Department of Electrical and Computer Engineer-
ing at Clemson University. Her research interests
include distributed computer systems and com-
puter networks, with an emphasis on content
delivery networks, mobile computing, wireless
sensor networks, and cloud. She is a Microsoft Faculty Fellow of 2010, a
senior member of the IEEE, and a member of the ACM.
Ze Li received the BS degree in electronics and
information engineering from the Huazhong Uni-
versity of Science and Technology, China, in
2007, and the PhD degree from the Department
of Electrical and Computer Engineering of Clem-
son University, in 2012. His research interests
include distributed networks, with an emphasis
on content delivery networks, wireless multi-hop
cellular networks, game theory, and data mining.
Chenxi Qiu received the BS degree in telecom-
munication engineering from Xidian University,
China, in 2009. He currently working toward the
PhD degree in the Department of Electrical and
Computer Engineering at Clemson University,
SC. His research interests include sensor net-
works and wireless networks.
For more information on this or any other computing topic,
please visit our Digital Library at www.computer.org/publications/dlib.
8143271457
8143271457

A distributed three hop routing protocol to increase the capacity of hybrid wireless networks

More Related Content

What's hot (20)

Similar to A distributed three hop routing protocol to increase the capacity of hybrid wireless networks (20)

More from Pvrtechnologies Nellore (20)

Recently uploaded (20)

A distributed three hop routing protocol to increase the capacity of hybrid wireless networks