Give

Graduate Exam Abstract


Faeze Heydaryanfroshani

Ph.D. Preliminary
February 13, 2019, 12:00 pm - 2:00 pm
LSC Room 226
Medium access control for distributed systems

Abstract: Due to the dramatic increasing
demand for wireless communication
and limited available resources such
as bandwidth, efficient access to the
shared wireless channels among
wireless users has become a critical
issue. Classical information theory and
classical network theory have been
exploring this issue from different
perspectives. Classical information
theory provides guidance to design
communication systems by addressing
performance limitations of the system
such as channel capacity. The
assumption in deriving performance
limitation of the system is that users
can coordinate with each other to
choose the best communication
parameters in terms of optimizing a
utility function. This type of
communication is called “coordinated
communication”. Since classical
information theory was developed for
applications with long communication
messages and small number of users,
overhead of coordinating users can be
ignored.
Most of current communication
networks contain large number of
devices with various types of
applications. To establish such huge
and complex networks modularization
is necessary. Classical network theory
has suggested the layering
architecture such as the Open System
Interface (OSI) model, to achieve
modularization in communication
networks. The layering architecture
organizes networking functions into
abstraction layers with clearly defined
interfaces. This allows network design
and optimization be limited to one or a
small number of neighboring layers
without the worry of whether the
results can fit into the network system.
Classical network theory mainly
focuses on connecting a large number
of users to support a diversified range
of network functions. Communication
efficiency of the network is often a
secondary concern.
Nowadays significant proportion of
wireless messages are transmitted
using distributed protocols where each
user chooses its communication
parameters individually. Due to bursty
traffic and highly dynamic users’
activities, coordinating a large number
of users in wireless communication
can be infeasible or expensive in terms
of overhead. Therefore, classical
information theory, which assumes full
users coordination, cannot be easily
extended to distributed networks. On
the other hand, classical network
theory does not provide effective
support for efficient distributed
communication and networking. In
current network architecture, data link
user only decides whether a packet
should be transmitted or not. Other
communication details are determined
at physical layer. In distributed
networks when optimization cannot be
done fully at the physical layer, data
link layer should get involved into
communication adaptation. However,
with binary transmission options,
advanced wireless capabilities such as
power and rate adaptation become
irrelevant at the data link layer. This
motivated our proposal of an
enhanced physical-link layer interface.
In the enhanced physical-link layer
interface, data link layer user is
equipped with multiple transmission
options plus an idling option. Each
option represents a particular
combination of communication
parameters such as communication
rate and transmission power. New
channel coding proposed in [] for
distributed communication, supports
enhanced interface by providing an
ensemble of channel codes at the
physical layer, each of them
corresponding to a link layer
transmission option. Advanced
wireless adaptation can be
implemented using enhanced
physical-link layer interface to optimize
a general utility function. Based on
decision made at data link layer,
physical layer chooses one code to
send its message. However, to
preserve modularity in the network
architectures, we assume data link
layer is constrained to choose one
transmission option from the provided
options to transmit each message. The
coding choice may be shared neither
with other transmitters nor with the
receiver. Because users are not
coordinated, collision may happen at
the receiver and reliable message
recovery may not always be possible.
This new channel coding enables
collision detection by defining an
achievable region on communication
rates such that packet recovery is
supported for rates inside the region
and collision detection is supported for
rates outside the region.
In current network architecture, each
user is equipped with single
transmission option. When collision
happens, users involved in collision
back off by decreasing their
transmission probabilities. From well-
known results in channel coding, we
know that, it would be more efficient to
decrease communication rate in
response to packet collision rather
than decreasing transmission
probability. When each user is
equipped with multiple transmission
options, each option corresponding to
a particular combination of rate and
power, how system should respond to
success transmission and packet
collision and how to support such
functions at the physical layer are
questions that should be answered.
We propose a distributed MAC
algorithm to optimize a general utility
function with/without enhanced
interface. We consider a distributed
wireless multiple access network with
unknown number of homogenous
users, each user is backlogged with a
saturated message queue. We
assume a general communication
channel rather than a collision channel
considered by most of existing MAC
algorithms. At first we consider
multiple access network with identical
users, then we extend it to the
networks with two different groups of
users. In distributed wireless networks,
users often access the channel
opportunistically which makes it
difficult to know the number of active
users at each time. Although
significant proportion of messages in
distributed networks are bursty short
messages, the assumption of
saturated message queue is to simplify
problem formulation. Bursty message
arrival leads to coupling between
users’ communication activities and
the statistical properties of such
coupling is often difficult to understand.
Furthermore, analyzing the system
with the assumption of saturated
message queue provides an
achievable bound on system
performance for network performance
with bursty message arrival.


Adviser: Dr. Rockey Luo
Co-Adviser: NA
Non-ECE Member: Dr. Haonan Wang, Statistics
Member 3: Dr. Ali Pezeshki, ECE
Addional Members: Dr. Liuqing Yang, ECE

Publications:
Y. Tang, F. Heydaryan, J. Luo, "Distributed Coding in A Multiple Access Environment," Foundations and Trends in Networking, Vol. 12, No. 4, pp. 260-412, Now Publishers, 2018.

Y. Tang, F. Heydaryan, J. Luo, "Distributed Multiple Access with A General Link Layer Channel," Ad Hoc Networks, Vol. 85, pp. 120-130, March 2019.

Y. Tang, F. Heydaryan, J. Luo, "Distributed Capacity of A Multiple Access Channel," IEEE Wireless Communications and Networking Conference, Marakech, Morocco, April 2019.

Y. Tang, F. Heydaryan, J. Luo, "On Utility Optimization in Distributed Multiple Access over a Multi-packet Reception Channel," IEEE International Symposium on Information Theory, Barcelona, Spain, July 2016.


Program of Study:
ECE651
ECE516
ECE514
ECE658
ECE652
STAT720
CS545
GRAD544