which essentially requires the convergence of two different IT technologies. This

Chapter 1. Introduction

1.1 Background

In recent years, the rapid growth of digital wireless telephony services and the

Internet access raises a tremendous demand for both services merged in one device,

which essentially requires the convergence of two different IT technologies. This

phenomenon will also influence people on the way of using network services. Besides

voice services, the data services, such as e-mail retrieval, web browsing and tickets

ordering, are thought to be available for access everywhere and every moment.

Although both data service and voice services can be delivered over the same air

interface, the data services have three major differences from voice services: i).

Asymmetry: The data services have much higher forward link (downlink) data traffic

than reverse link (uplink) data traffic. On the other hand, the voice services basically

have equal traffic on uplink and downlink. ii) Delay tolerance: Most data services can

tolerate a few seconds to several hours delay (eg. downloading a large file through FTP

protocol) while voice services can only tolerate about 200 milliseconds. iii) Burstiness:

When there are data to be sent, they should be sent as fast as possible, which

necessitates consuming the whole bandwidth. In voice services, the traffic generally

occupies a small fixed portion of available bandwidth. These differences have raised

the issue of how the data can be transmitted efficiently.

In wireless environment, radio spectrum is a scarce resource. In order to make

good use of radio spectrum, it is natural to use a cellular communication system. In

such systems, the basic idea is that we divide the intended coverage areas into smaller

cells. In each cell, there is one base station (BS) responsible for communicating with

mobile terminals (MTs) in that cell. The MTs are the devices one uses to download

data or communicate with other users. In this way, the radio spectrum used in one cell

can also be used in other cells.

How well the traffic can be transmitted depends on bandwidth, channel

conditions and transmission methods. Bandwidth is usually fixed, which is therefore

beyond our consideration. In a cellular system, the channel conditions depend on

several factors: i) Path loss, which includes distance-related attenuation as well as

slow fading and fast fading due to environment change. ii) Interference, which

includes co-channel interference from other BSs and MTs, adjacent channel

interference, and background noise. iii) Multi-path delay spread. iv) Doppler effects

due to mobility.

Transmission methods include any means we manage the resources. For example,

how the traffic is encoded at the physical layer, the media access control methods, the

power control methods, the packet scheduling algorithms, …., etc. Some of the

methods can use the resources efficiently, while some of them may improve channel

condition.

A sophisticated modern wireless device usually supports various transmission

rates. How it determines the rate is as follows. The receiver first measures the channel

state, in terms of SINR (signal to noise plus interference ratio), determines what data

rate it can receive, and tells this information to the transmitter. The transmitter then

encodes the data by using an encoding algorithm suitable for the designated

transmitting rate and transmits the data. The receiver then receives and decodes the

data.

1.2 Objective

The objective of this paper is to improve the forward link data services. The

methods for the purpose include packet scheduling and power control. In particular, we

consider the proportional fairness algorithm as a baseline policy for opportunistic

scheduling. The spirit of such scheduling is as flows:

The channel condition varies with time, so a user (i.e. a MT) may be in a better

next moment. If we can transmit a packet to a user when he is in a relatively good

channel condition, the transmission rate can be higher, which thus reduces the needed

air time. This is essentially one way of opportunistic scheduling for exploiting

multiuser diversity.

We assume the system is a time sharing (TDM) system where a BS transmits

data to only one user at a time. Thus the system suffers no intra-cell interference. To

reduce inter-cell interference, we can reduce the transmission power of neighboring

BSs. But for the users in neighboring cells, this also reduces their SINR. Thus, we

have to determine the right time when the transmission power should be reduced. For

the purpose we will propose a power control scheme.

We will use Monte Carlo simulation to show the performance of our proposed

algorithms.

1.3 Related Work

As for packet scheduling, the most well known and widely used one is

round-robin which might also be the simplest method among all other methods. By the

scheme, all users take turns to be served. Every user is served for every fixed time

interval and is served for the same portion of air time. One disadvantage of

round-robin is that it can’t exploit multiuser diversity effect.

A number of packet scheduling algorithms have been discussed in the literature,

which can be divided into three classes. i) Moving window: The concepts of these

methods are using a sliding window and divide the window into several time slices. It

needs to use an algorithm to determine which time slices to serve which user. ii)

Earliest deadline first: These kinds of methods are to determine the remaining service

time of a user (or a packet). The shorter the remaining service time is, the higher the

priority of the user will be. iii) Generalized processor sharing approaches (GPS) [2]: a

GPS server is work conserving and guarantees that every user has a pre-assigned

proportion of resources. An example algorithm of this class is QUALCOMM’s

proportional fair algorithm (Q-PFA) [1], [10] which guarantees every user being

served for approximately equal air time and has higher capacity over round-robin. The

algorithm is what we focus on and will be described in more details in later chapter. In

[17], the authors present an opportunistic transmission scheduling policy. This policy

uses two vectors, U and v. U is the performance vector that can be the measurement

result of throughput, SINR, monetary value, etc. v is the parameters continually

updated by its update mechanism. This policy chooses the user i with maximum of (U

+ v

).

The concept of using a power control method to increase the system capacity has

emerged for only a few years. In [11], the authors use multiple transmit antennas to

induce large and fast channel fluctuations so that the system can exploit more on

multiuser diversity. In [5], the simple power control scheme (SPC) is to divide all BSs

into two groups and sets the emitting power of one group to high and another group to

low. Then, their power level switches between high and low alternatively. In [7], the

authors proposed a “search-and-lock” algorithm to search the lower-load cells, set their

power to low and “lock” their adjacent cells to high. They proposed two adaptive

schemes, namely adaptive power level control (AP) and adaptive power level control

with hysteresis approach (APH). AP uses one threshold of loading factor as a

condition for switching power level, while APH uses two thresholds to avoid frequent

power levels switching.

1.4 Thesis Organization

The remaining of this thesis is organized as follows. In the next chapter, we

present our system model. In chapter 3, we describe Q-PFA and discuss our packet

scheduling policies. In chapter 4, we describe SPC and our power control schemes.

Chapter 5 presents simulation model and experiment results. Finally, we give our

concluding remark and some points of future work in chapter 6.