detected at the receiver side. Taking advantage of these
clean subcarriers, we can record each attached signal in a
training sequence for the purpose of attachment cancelation
and data recovery in subsequent payload data packet.
Specifically, the received signal with Attachments on clean
subcarriers of a training sequence can be expressed as
y
clean
t½¼y
attach
t½þnt½: ð1Þ
Accordingly, the received signal in subsequent data
symbol with both data and attached signals can be
expressed as
y
raw
t½¼y
data
t½þy
attach
t½þnt½; ð2Þ
where y
attach
½t¼H Attach½t and y
data
½t¼H Data½t
are attached signals and data signals, respectively, after
traversing channels to the receiver. H refers to the
corresponding channel impulse response which can be
calculated using a training sequence and n½t refers to a
random complex noise. Therefore, the original data signal
can be recovered by canceling the attached signal from the
received signal in data symbol, that is
Data
i
t½¼
y
raw
i
t½y
clean
i
t½
H
: ð3Þ
Fig. 3 illustrates Attachment Coding in time/frequency
domain. A training sequence carries the recording attached
signals for cancelation and subsequent data packet carries the
actually coordination signals. The feasibility of Attachment
Coding is analyzed in the supplemental files, available
online, serving as design principle for Attachment Coding.
3 AR-MAC PROTOCOL
To demonstrate the effectiveness of Attachment Coding, we
present a cross-layer design, AR-MAC. AR-MAC builds on
top of Attachment Coding, which aims to solve the exposed
terminal problem in distributed wireless networks. First, we
give an overview and design challenges of AR-MAC.
Detailed modules are then presented to see how we address
these challenges. Finally, we talk about some points related
to an AR-MAC design.
3.1 AR-MAC Overview
To identify whether a node is exposed terminal or not in a
distributed network, we need two kinds of information:
1) the ongoing sender-receiver pairs; and 2) the neighbor-
hood information within two-hop collision domain. As
illustrated in Fig. 4, nodes can be characterized into two
types: current sender-receiver pairs and intended sender-
receiver pairs. Suppose an intended sender (IS) Bob has
some packets to transmit, he should guarantee that his
intended receiver Alice or Lucy is available and other
current receivers (CRs) like Coral will not be interfered by
his transmission.
Based on the above observation, we propose AR-MAC,
which utilizes Attachment Coding and neighborhood list to
provide the above two information, including the ongoing
transmission on air, and the two-hop neighborhood list.
Without loss of generality, we assume that the interference
and reception range are equal. To obtain the information of
current transmissions, each sender modulates its transmis-
sion information into Attachments (A-RTS) (like who are the
sender and receiver for this transmission). To obtain the
information of two-hop neighborhood, nodes can periodi-
cally be broadcasting their one-hop neighborhood list in
their vicinity [3], either broadcasting stand-alone packets for
the list, or piggybacking the list with routing beacons.
The design principle of AR-MAC is simple and efficient.
However, there remains several implementation challenges
when bring AR-MAC into practice:
. First, Attachment cannot carry too much information
due to its limited bandwidth. Thus, the format of A-
RTS should be designed efficiently.
. Second, distributed networks are always unsynchro-
nized with variable-length packet. Thus, it is difficult
for a node to obtain A-RTS whenever it needs.
. Last, any strategy that tries to utilize exposed
terminals has to handle ACK collision with other
data transmissions, and data transmissions can also
collide with themselves. These collisions should be
treated carefully to increase PRR.
To address these challenges, AR-MAC consists of two
stages: RTS sense and collision resolution. Every node first
goes through RTS sense to contend for channel access.
Afterward, collision resolution will handle collisions among
different kinds of transmissions.
3.2 RTS Sense
RTS sense initiates a normal contention for data transmis-
sion opportunity. Instead of carrier sense in CSMA that
detects carrier waves before trying to send, RTS sense
simply listens to A-RTS signals attached on carrier waves.
RTS sense contains two parts: RTS attachment and channel
access decision.
3.2.1 RTS Attachment
To modulate the current transmission information into A-
RTS, a specialized hash format is adopted. Each A-RTS
WANG ET AL.: ATTACHED-RTS: ELIMINATING AN EXPOSED TERMINAL PROBLEM IN WIRELESS NETWORKS 1291
Fig. 3. Illustrated example of attached control messages.
Fig. 4. Overview of Attached-RTS.