PhD Thesis Proposal Defence


"Packet Loss Analysis for Real-time Wireless Traffic through Delay Models"

By

Mr. Kai-Wing Lee


Abstract

Queueing models have been used successfully to analyze systems where
queue buffer overflow is the main concern. However, in systems with delay
constraints where actions need to be taken depending on the delay,
queueing models may not be appropriate since delay information is not
available and consequently approximations need to be made. In the proposed
work, a new type of model providing delay information, called the delay
model, is developed to facilitate the solving of this problem.

In the systems considered, wireless packets with delay constraints are
transmitted through an erroneous channel modeled by a two-state markov
chain. If a packet transmission is not successful, the packet is
retransmitted until a delay limit is exceeded. At that time the packet is
discarded and the transmission of the next packet begins. This packet
dropping process has a significant impact on packet loss probability but
is seldom considered in existing Markov models, probably due to the
difficulty of including delay information in the models.

To see if the concept of delay models really works, an attempt to solve a
problem with correlated errors and geometric inter-arrival times has been
done. The problem, which was described by a professor as "hard to
analyze", is solved and a closed-form expression for packet loss
probability is derived. Apart from making delay information available, the
delay model approach also has the advantage of capable to include a simple
arrival process in the delay dimension of the Markov chain. This is
illustrated in a simple model with correlated arrivals, which has a single
dimension only. Minimizing the number of dimensions in Markov models is
important since it can reduce the algebraic manipulations required to
obtain closed-form solutions. Usually Markov models with more than a few
dimensions are difficult to solve. Since both correlated errors and
correlated arrivals have significant impact on packet loss probability, it
is desirable to solve a system with both of them and surprisingly the
complexity of the expression for packet loss probability does not increase
substantially. Further generalization seems to be difficult.

Since different priorities for different wireless terminals (WTs) may
exist in the real world, priority systems are considered. They include a
simple one giving twice the bandwidth to high priority WTs and a second
one with a more sophisticated threshold-based extra slot assignment
scheme.


Date:				Thursday, 22 May 2003

Time:				3:00p.m.-5:00p.m.

Venue:				Room 4333
				Lift 3

Committee Members:			Prof. Samuel Chanson (Supervisor)
				Dr. Bo Li (Chairman)
				Dr. Brahim Bensaou
				Dr. Jogesh Muppala


**** ALL are Welcome ****