17-12-2012, 05:43 PM
CROSS-LAYER DESIGN OF AD HOC NETWORKS FOR REAL-TIME VIDEO STREAMING
1CROSS-LAYER DESIGN.pdf (Size: 171.84 KB / Downloads: 19)
ABSTRACT
Cross-layer design breaks away from traditional
network design where each layer of the
protocol stack operates independently. We
explore the potential synergies of exchanging
information between different layers to support
real-time video streaming. In this new approach
information is exchanged between different layers
of the protocol stack, and end-to-end performance
is optimized by adapting to this
information at each protocol layer. We discuss
key parameters used in the cross-layer information
exchange along with the associated crosslayer
adaptation. Substantial performance gains
through this cross-layer design are demonstrated
for video streaming.
INTRODUCTION
An ad hoc wireless network is a collection of
wireless nodes that self-configure to form a network
without the aid of any established infrastructure.
Some or possibly all of these nodes
are mobile. These networks are extremely compelling
for applications where a communications
infrastructure is too expensive to deploy, cannot
be deployed quickly, or is simply not feasible.
There are numerous potential applications for
ad hoc wireless networks, ranging from multihop
wireless broadband Internet access, to sensor
networks, to building or highway automation, to
voice, image, and video communication for disaster
areas.
The lack of established infrastructure, the
network and channel dynamics, and the nature
of the wireless medium offer an unprecedented
set of challenges in supporting demanding applications
over ad hoc wireless networks. The wireless
channel is inherently a broadcast medium,
so transmissions from different nodes interfere
with each other. The quality of wireless links
vary over time and space due to interference,
multipath fading, and shadowing. Network conditions
are highly dynamic as nodes join and
leave the network in an unpredictable manner.
RELATED WORK
Supporting multimedia applications over wireless
links has been one of the main fields of attention
in the networking and video coding communities
in the last decade. For such applications, it is
well known that the separation principle of
source and channel coding put forth in Shannon’s
information theoretic framework does not
hold. Hence, joint source and channel coding
techniques have been proposed to overcome the
challenges of wireless links. However, these have
not yet led to a unified solution to this problem,
as explained in [1]. Nevertheless, these considerations
have strongly influenced the video community
in the design of the new H.264 video
coding standard, which incorporates a highly
flexible syntax well suited to network transmission
[2].
As the number of nodes of a wireless network
grows, interference increases, reducing the
achievable data rates. In a landmark paper, the
capacity of a static wireless ad hoc network is
shown to asymptotically vanish as the number of
users increases [3]. However, recent results show
that in more practical settings, the high data
rates and low delay constraints of multimedia
applications may be supported [4, 5]. The 802.11
protocol operating in ad hoc mode provides an
interesting benchmark against which proposed
designs for ad hoc networks may compete. In its
simplest form, it allows for one user at a time to
be active within a carrier sense region (typically
on the order of a few hundred meters). As only
active hosts occupy the wireless medium, this
medium access control (MAC) protocol comes
closer to the achievable capacity than time-division
multiple access (TDMA) or frequency-division
multiple access (FDMA) systems. In this
sense it already incorporates some cross-layering.
However, this design is restrictive in more
than one way and may have to be surpassed to
enable efficient media streaming.
CROSS-LAYER DESIGN FRAMEWORK
A cross-layer approach to network design seeks
to enhance the performance of a system by jointly
designing multiple protocol layers. This
approach allows upper layers to better adapt
their strategies to varying link and network conditions.
The resulting flexibility helps to improve
end-to-end performance given network resources
and dynamics. These design concepts are particularly
useful for supporting delay-constrained
applications such as video.
A cross-layer approach to network design can
significantly increase the design complexity.
Indeed, protocol layers are extremely useful in
allowing designers to optimize a single protocol
layer design without the complexity and expertise
associated with considering other layers.
Thus, cross-layer design should not eliminate the
design advantages of layering. Keeping some
form of separation, while allowing layers to
actively interact, appears to be a good compromise
for enabling interaction between layers
without eliminating the layering principle. In
such a structure each layer is characterized by
some key parameters, which are passed to the
adjacent layers to help them determine the operation
modes that will best suit the current channel,
network, and application conditions. In such
a design each layer is not oblivious of the other
layers, but interacts with them to find its optimal
operational point.
ADAPTIVE LINK LAYER TECHNIQUE
Wireless link throughput is severely affected by
channel impairments such as shadowing, multipath
fading, and interference. Adaptive modulation
is an efficient technique to improve the data
rate by adapting link layer design variables to
the variations of the channel environment. These
parameters may include modulation, coding,
transmitter power, target BER, symbol rate, and
combinations of these parameters.
In general, adaptive link layer techniques
have not been considered in a cross-layer framework.
In this section we consider two adaptive
link layer techniques to improve link throughput.
We first consider adapting the packet length,
given the current SINR and link layer parameters,
to optimize throughput. In addition, for a
fixed packet length, we consider optimizing link
layer parameters such as symbol rate and constellation
size for maximal throughput. Indeed,
when the packet length is too large, packet error
rate (PER) increases, and throughput is limited
by frequent retransmissions.
SCHEDULING AND RATE ALLOCATION
In this section we describe the additional
improvement that can be obtained through the
use of smart packet scheduling, and discuss how
to determine the optimal operating rate for
streaming.
CONGESTION-DISTORTION OPTIMIZED SCHEDULING
Losses on the wireless medium are inevitable
due to interference, collisions, and mobility.
The absence of a packet at the decoder causes
a decoding error, which translates into a quality
drop that may propagate to subsequent
frames. This effect is illustrated in the sudden
quality drops present in Fig. 4, mainly caused
by losses on broken links. These drops may be
mitigated by smart scheduling at the transport
layer.
Traditional transport layer protocols such as
TCP provide reliable transmission but are
unaware of delay requirements and relative
importance of packets. In cross-layer design
superfluous transmissions may be avoided by
taking into account application layer delay constraints.
In [11] an even more advanced technique
is proposed, which seeks optimal
transmission schedules based on the importance
of each packet of a video stream. This type of
scheduling aims at maximizing the decoded
video quality at the receiver while abiding with
a rate constraint. In congestion-limited situations,
it is beneficial to use instead a congestion-
distortion optimized (CoDiO) scheduler,
which limits end-to-end delay [12]. This metric
better reflects the impact of a user’s transmissions
on the congestion of a network. In addition,
is inherently adaptive to time-varying
network conditions.
CONCLUSIONS
The unique characteristics of wireless ad hoc
networks call for new design paradigms that
move beyond conventional layering. While
joint optimization allows interaction and flexible
resource allocation across the network pronFigure
tocol stack, the growing complexity may be
prohibitive for practical implementation. To
strike a balance between performance gains
and design complexity, it is important to keep
the abstraction of layering while allowing
information exchange between adjacent layers.
The large gains foreseen by enabling different
layers to collaborate are promising for
demanding applications such as audio-visual
conversations.