CHAPTER 1

INTRODUCTION

1.1 PRELUDE

The inventions of telegraph and telephone during the 19' century were the

milestones in the history of telecommunications. Till then, writing was the most

predominant means of communication. At the beginning of the 20' century,

Guglielmo Marconi transmitted the first transatlantic radio signal. Radio

communications, since then have continually improved. It was during the Second

World War, that wireless technologies underwent a remarkable improvement.

Further, the development of cellular concept by Bell labs in 1970's fueled the

growth of wireless networks globally and made it much more pervasive than

anyone could have really imagined. Newer wireless systems and standards started

to evolve due to the wide spread success of the cellular concept, which paved the

way for the growth of wireless communication networks. This has resulted in an

explosive outburst of wireless subscriber population and newer mobile services.

1.1.1 Evolution of Cellular Systems and Standards

It has become very common to categorize the advancements in wireless

technologies by a generation label. The first generation (1G) mobile telephony

systems, which began in 1980's, were analog frequency division multiple access

(FDMA) systems designed for supporting voice services. In FDMA systems, the

frequency band is divided into channels of equal bandwidth, such that every user

is assigned a different frequency slot to avoid interference (Fig.l.1). A full-duplex

FDMA transmission requires two channels, one for transmitting and the other for

receiving. Advanced mobile phone system (AMPS), nordic mobile systems

NMS) and total access communication system (TACS) belong to this generation.

FDM A TDMA CDM A (Frequency Division (Time Dibision (Code Division

Multiple Access) Multiple Access) Multiple Access)

Fig.l.1 Various multiple access techniques

Major problemr u ~ t h them are lirn~ted cobcrage. poor quality and capacity. [n

addition lo this, dernand for better terlices such as unrestricted roarn~ng, data and

supple~nen[ar) scrlices uere the rnaln cause for probiders to find a better system.

Second peneratlon i?G) mobile telephone networks were the log~cal next

stage In the de\rlopment of wireless systernr to 1G. For the first time, the 2G

systems ~ntroduced mob~le phones u ~ t h purely digital technology. Although many

of the principles in\ol\ed in 2G system are same as that of IG. the former was

capable of providlnp more adlanced feature such as caller ident~ty and text

messaging.

Tlrne d l ~ ~ s l o n multiple access (TDMA) is a ?G uireless technology that

p r o ~ ~ d e s c~rcuit-suitchcd d a ~ a services along uith a high-quality \oice. The most

popular 2G standards arc global s)stem for mobile comrrlunications (GSM). :I

European bawd sbstem and Interim standard-136 (IS-136) fo l lo~ed in North

America. In these systems, the indnidual mobile users are cyclically assigned a

frequent) band for exclus~ve use on14 for the duration of a time slot (Fig.1 I ) The

connection is retained until it 1s termmated or handed off to another cell. No other

user using the same frequency band 1s assigned the same time slot and thus users

within a cell do not interfere. The capabilit) of TDMA systems to effectively

utilize the spectrum makes it flexible to handle multiple bit rates and thereby,

prokide three times more capacity than AMPS

GSM supports eight time slotted users for each 200 kHz radio channel

with data rates of 13 kbps, unlike IS-136, which supports three time sloned users

for each 30 kHz radio channel. Short messaging service (SMS) became a very

popular feature of GSM. In the US, using the code division multiple access

(CDMA) technology, Qualcomm has launched an alternative cellular standard

namely, interim standard - 95 (IS-95). Every user is distinguished using different

spread codes in CDMA systems (Fig.] .I). Direct sequence CDMA @S-CDMA)

and frequency hop CDMA (FH-CDMA) are typically used forms ofmodulation in

these systems. One of the attractive features of CDMA technology is the universal

frequency reuse, in which, all users occupy a common frequency allocation. This

increases the spectrum usage and eliminates the need to plan different frequency

allocations for neighbouring users or cells. Thus, it is more bandwidth efficient

than FDMA or TDMA systems. For instance, IS-95 @S-CDMA system) supports

upto 64 users for simultaneous transmissions on each 1.25 MHz band by assigning

onhogonal codes to avoid interference.

The data rates in 2G were limited due to the use of circuit switched

techniques. Although these systems did overcome all the shortcomings of the IG,

the need for high-speed data prompted further enhancement of data services. In an

effon to retrofit the 2G standards to improve its compatibility with increased data

rates, new data centric standards that can be overlaid on 2G standards have been

developed. This new generation termed as 2.5G, has been mainly intended to

provide high rate data traffic with better quality and higher capacity at affordable

prices to data users.

High speed circuit switched data (HSCSD) was an initial attempt to

provide enhancement of data services upto 28.8 kbps for all current GSM

networks. Using this, subscribers can access data services three times faster. A

packet based data network named general packet mdio service (GPRS), well

suited for non-real time data services with a maximum support of 171.2 kbps

emerged as an alternative solution. GPRS supports multi-user sharing of radio

channels and time slots resulting in a higher capacity than that of HSCSD

networks. Subsequently, enhanced data rate for GSM evolution (EDGE), another

high speed data standard was evolved. It is capable of providing high data rates

3

upto 384 kbps using multiple modulation formats. IS-9SB, an interim data

solution for 2.5 CDMA provides both high speed packet and circuit switched data

access on a common CDMA radio channel at data rates upto 14.4 kbps besides

voice senices. The 2.5G solution suppons medium data rate services at the rate of

115.2 kbps using the code aggregation techniques. Low efficiency in handling

packet oriented services, low speed data transmissions and multiple standards that

prohibit roaming between regions fueled the growth of third generation (3G)

networks.

3G mobile telephone networks, also referred to as universal mobile

telecommunication systems (LiMTS) are the latest and recently deployed cellular

networks that work with packet switched techniques. The main 3G technologies

include wideband CDMA (W-CDMA) and CDMA 2000. Europe has settled on

the UMTS with W-CDMA as its chosen approach, while CDMA 2000 (upgraded

version of IS-95) has been adopted by the United States. The data rates supported

by the basic 2G networks were only 9.6 kbps. However, 3G systems are designed

to provide a range of data rates upto 144 kbps for moving vehicles, 384 kbps for

pedestrians and 2 Mbps for indoor or stationary users depending on the user's

speed and environment. Higher data transmission rates and increased capacity are

the significant features of 3G systems, which enable them to satisfy the increasing

demand for high speed data applications.

The next quantum leap to support unimaginable data rates will eventually

lead to fourth generation networks (4G). It is envisioned to do something that 3G

evolutions fail to do. 4G might come into reality very soon as the demand for

higher data rates continues everywhere as a future trend.

The increasing use of internet protocol (IF') based applications and

emergence of attractive multimedia terminals into the markets have made the

emerging wireless systems to follow the trend of increased data rates and

improved spectnun utility. To live upto these expectations, many technical

challenges have to be faced. The weliable nature of wireless links complicates

the task of managing the connections. Further, the wide variety of wireless

applications expects diverse quality of service (QoS) requirements from the

4

underlying network. This clearly emphasizes the need for mechanisms, which are

capable of managing diverse applications in the midst of highly unreliable

wireless links. Added to this, the potential need for wider spectrum or bandwidths

to provide high data rates in cellular evolution raises further questions on the

spectrum needs. The radio spectrum available for wireless data services and

systems is extremely scarce, while the demand for these services is growing at a

rapid pace. An engineering solution to meet the spectrum needs is to improve the

spectral efficiency.

These discussions clearly imply that adaptive solutions, which can govern

both quality guarantees and spectrum usage along w~th the capability of exploiting

the time varying nature of wireless links, will be more appropriate for the growing

demand on high speed data applications. Therefore, there is a need to focus on

various radio resource management (RRM) tools and explore avenues to address

the above specified needs.

I .2 RESEARCH APPROACH

QoS provisioning actually refers to a network's capability to provide

different levels of service to varying classes of traffic. It relies greatly on the radio

resource components like call admission controller, scheduler, power control etc,

which reflect the network's availability and transmission quality. Thus, the main

objective of the RRM techniques is to provide the necessary QoS guarantees

besides utilizing the available radio resources efficiently. As the quality of

wireless channel varies randomly with time, any anempt to provide determilustic

QoS (i.e., requiring zero QoS violation probability) will most likely result in

extremely conservative guarantees in cellular networks. This conservative

guarantee is absolutely useless, as the deterministically guaranteed quality in

fading channels is zero. Therefore, a greater emphasis has been laid on statistical

QoS in the present work.

The current proliferation of application-oriented mobile networks is

inefficient expensive and unlikely to satisfy long-term diverse quality

5

requirements. Thus, it is universally agreed that a richer set of QoS levels are still

needed to support emerging wireless applications. Further, the forthcoming

generations including 3G systems are expected to use packet-switching

technology, which is more efficient and faster than the traditional circuit-switched

systems. According to the world mobile markef currently about 75% of the world

wireless population still relies on TDMA mobiles, while the remaining have

switched over to CDMA mobiles. The efforts have thus been to enhance QoS in

both TDMA and CDMA transmission technologies. The ability of a

network in establishing and maintaining wireless connections provides the

measure of offered QoS level. Therefore, in the present study, the prime focus is

given to QoS from the network perspective.

Establishing and maintaining the connections are the two critical stages

encountered by any mobile user in accessing the cellular network. During the

connection establishment stage, the mobile user attempts to obtain the resource

necessary for the link quality. This task (to accept or to reject) is taken up by the

call admission controller, which is housed at each cell for making admission

decisions. The nature of the resource allocated depends on the type of

transmission technology. Once it is successful in getting the required resource, the

mobile user would start to transmit information. Due to user mobility, frequent

handoffs may occur during the connection lifetime. Sufficient resources have to be

allocated in all the handoff cells to ensure connection continuity. If it is not

available, the connection is terminated abruptly during the active lirk period,

which annoys the user to a greater extent. Therefore, one of the primary goals of

the network designer is to keep the probability of such an occurrence termed as

handoff dropping probability as low as possible. Thus, call admission

control (CAC) schemes are usually designed to limit the number of

new calls admined into the network, to felicitate the already admined calls to

obtain their desired QoS.

During the connection maintenance stage, no new usen enter into the

system. However, due to user mobility, each user will have to exchange their

s h e of resources. Throughout the connection lifetime, the scheduler is actively

involved in scheduling the resources among different usen according to the QoS

6

demands. It is essential to note that mobile users experience a time varying

performance (e.g., throughputs) due to time varying channel conditions.

Consequently, the scheduler should determine the next packet for transmission

based on the user channel or link conditions. However, this significantly affects

the user fairness (attention paid by the network towards the individuals) and

throughputs. Hence, good link adaptive scheduling schemes that achieve higher

resource utilization and at the same time improve user throughputs still remains to

be addressed.

The need for pushing the spechum to the extreme necessitates

enhancements in spectral efficiency, which is defined to be the number of bits

tmnsmined per symbol or equivalently expressed in terms of houghputs

@psMz). Therefore, in addition to the schedulers, suitable link adaptation

mechanisms like adaptive modulation and power control (PC) are also essential

during the connection span. Link adaptation aims at both quality and effective

utilization of the allotted resources by adapting the transmission parameters such

as transmission power, modulation andlor coding etc. Power control aims to

regulate the transmission power in such a way that the co-channel interference is

minimized. In order to avoid near far effects, minimum nansmission powers are

assigned to strong interfering mobiles. However, this kind of power distributions

limits the data rates as well. This problem can be solved to a great extent, if both

power control and adaptive modulation schemes are jointly executed. Focusing at

various resource management techniques separately may not yield appreciable

results. A substantial improvement in throughputs and shon-term quality needs

can be realized only when the scheduler work in unison with the other link

adaptation mechanisms. This possibility has been explored in the present study.

13 LITERATURE SURVEY

The present work deals with two major research issues: (i) Limiting both

handoff dropping and new call blocking rates during the connection establishment

phase using CAC algorithms. (ii) Ensuring QoS during the connection lifetime

with the help of schedulen and l i i adaptation mechanisms. An exhaustive

literature related to the present work has been presented.

In the past, several solutions have been proposed to reduce the radio

blocking probability due to new calls and handoffs for cellular systems. As the

handoff droppings are more serious than new call blockings, various solutions,

which could treat new and handoff calls differently have evolved [I-41. Handoff

prioritization by resewing a fixed number of guard channels at each

base station (BS) was one of the common solutions [I] . Queuing of hand over

requests was another possible option as a finite time interval is available for the

mobile or mobile station (MS) to spend in the handover area (21. While most of

the approaches aims at lowering handoff drops, Tekinay el a/ and Chia have

suggested a different solution to provide high radio spectrum utilisation by

reducing the new call dropping rates [3,4]. The obvious limitation of this solution

is the increased handoff dropping rates. which is an essential QoS metric.

Subsequently. Ramjee el a1 have introduced a fractional guard channel

policy, in which new calls were admined conditionally with some probability

instead of complete denial of service [ 5 ] . Although, it has shown some

improvement, the analysis was restricted only to a homogenous cellular network

uith a single cell in isolation. An altemative solution [6] referred to as distributed

CAC scheme to limit the handoff dropping probability within a threshold bound

was introduced. The call admissions are taken with the consent of the

neighbouring cells in a distributed manner. The performance analysis of this

scheme shows that it has not considered multiple services and non-uniform truff~c

conditions. Further, no measures have been taken to locate the exact handoff cells,

which the mobile users would probably visit. Later, the problems with the

distributed approach was addressed by Levine el al using the concept of a shadow

cluster. a set of cells to which an MS is likely to attach in the near future [7]. This

scheme assumes precise knowledge about individual mobile user's dynamics and

call holding patterns in the form of probability density functions. Such user

specitic information may not be available in all mobile units and it is not clear

how the scheme would handle such cases. Another alternative to distributed CAC

8

scheme was a cell-cluster based CAC suggested by Naghshineh et a1 [8]. The

number of mobile users admined into the cell-cluster is limited so that every

admitted user enjoys a predefined QoS. This is achieved through blind reservation

of radio resources in cells. Subsequently, a predictive and adaptive bandwidth

reservation scheme was evolved as the next step towards dynamic reservations

[9,10]. A knowledge base of previour cell history and future handoff cells for each

mobile user is the basic requirement of these schemes. In practice, mobile units

cannot provide these unrealistic inputs. A similar attempt was carried out with

fuzzy logic controller for a micro cell environment, where handoffs are frequent

[I I]. However, even the fuzzy logic controller could not address the above said

problem. These schemes have been trying to provide zero handoff dropping rates

with unrealistic assumptions.

Realising the importance of user mobility patterns, attention has been

directed towards predicting the user movements. Initially, this approach was tested

for indoor environments without considering the channel holding time [12,13].

Later, it has extended to cellular networks assuming that each mobile user will

handoff with a predeterm~ned mobile probdbility to any of its neighbouring cells

[14-161. Subsequently, an improved prediction mechanism based on user's

handoff time observations was proposed by Shen el a1 [ I 71. However, only a one-

dimensional cellular structure confined to single service is considered for the

study. This has prompted Olrveria el a1 to suggest no\el CAC mechanisms

suitable for real network, supporting multiple services [IS]. The proposed

algorithm adjusts the amount of reservations based on both handoff dropping and

bandwidth utilisation rates. Later, researchers 119-211 have extended these efforts

for adaptive multimedia services. Although these schemes seem to be attr;utive,

they have failed to perform well in the event of rapid changes in the number of

mobile users entering the cells within a short period.

Assumptions on user movement behaviour, blind reservations to reduce

the handoff dropping rates and using identical distribution for channel holding

times of both new calls and handoff calls are impractical. These unrealistic

assunlptions are the major limitations in most of the schemes discussed above.

However, these unrealistic assumptions have simplified the blocking analysis.

9

Unlike T D W D M A systems, in which admissions are limited to the

available number of channels, admissions in CDMA environment are limited by

the total interference power i n d u c e d by the mobile users. Therefore, the

capacity of the cell varies with the loadiig of the home and the neighbouring cells.

Thus, in CDMA systems, the CAC policies are highly correlated with power

control measures.

Initially, CAC algorithms designed for CDMA systems have completely

ignored the interference effects of the future handoff calls and admissions have

been made only based on the effective traffic load at the time of admission request

[22,23]. Several signal to interference ratio (SIR) based CAC schemes (24,251 for

DS-CDMA systems, incorporating the interference effects have been proposed

under the assumption that the base station receives the same signal power from

each of its mobile users. These schemes have also been found to be less efficient

in making admission decisions for time varying MIC scenario. This fueled

effective admission policies, referred to as interactive CAC to work in

dynamically changing t rf i lc scenario [26.27]. In these proposals, a new call

request is admitted only if it does not introduce excessive interference into the

system. Although interactive CAC is an ideal policy. it requires a longer

convergence period, which makes it impractical. Further, its effectiveness is tested

in a homogeneous environment. Later. Evans er a1 and Ho er a1 have attempted to

suggest admission procedures for multiple services [28,29]. In these proposals,

fixed admission thresholds are set to determine the number of admissible users.

The serious limitation of this scheme 1s its failure to take care of the time variant

behaviow of capacity and SIR constraints before admitting a mobile. In such

cases, there is no guarantee on quality support, which is essentid for

heterogeneous applications.

Subsequently, attempts have been made successfully on admission policies

suitable for heterogeneous networks [30,31]. However, they have modeled the

system using a single dimensional Markov chain by considering identical

assumptions for both new and handoff calls, though they are to be treated

separately. Kim er a/ and Redana et a1 have extended the analysis of system model

using a multidimensional approach [32,33]. In these admission procedures, the

I0

total interference level at the base station is adopted as a measure of load. The

persistent problem with all these schemes is the lack of importance for spechum

usage besides minimizing the call droppings. Consequently, with primary focus on

spectrum utility, a new concept of call degradation was introduced. Chlamiac e t a1

and Chou el a1 have presented an analytical model for combined degradation and

traffic restriction mechanism [34,35]. However, this has resulted in quality

degradation due to lack of mechanisms to predict the movement tracks. The user

movement tracks are not considered for most of the proposals due to the high

degree of variability associated with the CDMA capacity. Further, it was assumed

in most of the proposed analytical models that the interference power offered by

every call is same. These are the two major shortcomings reported in the

literature.

Most of the scheduling schemes for wireless networks are derived from

wire-line scheduling policies. The scheduling mechanisms, mostly reported in the

literature have evolved from generalized processor sharing, as it is an ideal

scheduling policy for wire line environments [36-401. Weighted round robin

(WRR) and deficit round robin (DRR), simplest approximations of the generalized

processor sharing scheduler are suitable for packet based networks 1391. They are

suitable only for fixed packet size and do not provide deadlines on delay. On the

other hand, fair scheduling is more suitable for such networks, as they provide

minimum bandwidth guarantee and bounded end-to-end delay :o all constraint

flows [36-381. Unlike wire-line systems, wireless systems suffer from high bit

error rate (BER) due to the location dependent and the random name of channel

effects. As suggested by Benett el al, channel state dependent scheduling and

compensation are the only solutions to improve the above situation [40].

Therefore, wireless fair scheduling algorithms employing such solutions

have started to emerge slowly [41-441. Each algorithm has its own specific

realization of compensation and thus its performance varies in a wide range in

terms of delay bound, throughput guarantee, short-term fairness and gracefulness

of degradation. However, these algorithms [41-431, degrade the performance of

leadilig flows during compensation. The main cause for this problem is that the

bandwidth for compensation cannot be reserved in advance. This is due to the

11

dynamic changes in resource requirements in accordance with the time varying

nature of the channel states. To resolve this problem, Ramanathan et a1 have

proposed a long term fairness server (LTFS) policy, where a portion of total

bandwidth is pre-allocated for compensation [44]. Therefore, there is no QoS

degradation during compensation. However, due anention was not given for

fairness among the flows. Recently, Jeong el a/ [45] have introduced two different

algorithms such as wireless general processor sharing (WGPS) and packetized

WGPS (PWGPS). They have provided compensation by increasing the service

share of the affected flows to a predetermined amount, These scheduling policies

assumed that they have a perfect control over the wireless channels. In practice,

suitable prediction mechanisms are required to provide channel state information

as inputs to the scheduler for making effective scheduling decisions.

Exhaustive research work on channel predictions reported in the literature

shows that initially analysis on signal estimation in fading channels was focused

on linear prediction techniques [46-481. Most of them have employed very

complex mathematical analytical tools for channel characterization. Further, in

some of the proposals. accuracy of the schemes greatly relies on the prediction

horizon. As an alternative solution to reduce these complex computations. several

heuristic based prediction algorithms have evolved and mostly they have been

evaluated jointly with scheduling mechanisms.

One such scheduler logic, referred to as channel dependent packet

scheduling has been proposed by Fragouli er a1 [49]. In this approach, channels

have been modeled as either good or bad. In reality, it is too simple to characterize

realistic wireless channels using binary models especially for data serrrices.

Songwu el a1 have presented a heuristic scheduler logic, based on the reception of

the last received packet [SO]. The channel state in the next time instant has been

assumed the same as the previous instant. The major drawback is that the

predictor will not observe the good state periods during which packets could have

been transmitted, if the actual fade periods are shorter than the assumed periods.

Later, a long range channel prediction algorithm to predict the future

fading attenuation based on the past channel samples has been introduced by

12

Hu el a1 [51]. The accuracy of this method is limited by the inaccuracies in the

estimation of the channel parameters. Subsequently, Sfernad el a1 have proposed a

prediction technique for fading channels with a different evaluation method [52].

The metric used for evaluation in this case is the percentage of time during which

the predicted signal is within a certain bound of the actual signal. Like previous

proposals, this scheme also has exhibited poor prediction accuracy, which

decreases with the prediction horizon. In summary, both heuristic and

mathematical methods are not perfect. The predicted results ale inaccurate and do

not correspond exactly to the future behaviour of the channel. The effect of this

inaccuracy on the performance of scheduling mechanism needs to be studied.

Further, all these schemes have been confined to homogeneous platforms.

This fueled a lot more attention towards multiservice scheduling policies.

A multi class priority fair queuing (MPFQ) that fairly shares the wireless link

bandwidth among different classes has been presented by Mnorman el a1 and

Ren Xiuli er a1 [53,54]. Due to fair sharing of resources, these approaches have

suffered from poor throughputs. Several scheduling solutions for achieving

optimal throughputs 155,561 have been proposed by Andrew el al. However, it is

based on the assumption that the queues are always maintained stable.

Several researchers have focused their attention on CDMA scheduling

policies for mult~ple services besides TDMA scheduling 155-591. These schemes

have attempted to guarantee long-term average throughput to each user. However,

high performance gains have been achieved only at the cost of additional

complexity and signaling overhead. Subsequently, Berggren el a1 have proposed

a joint power conuol and intracell scheduling in DS-CDMA systems [60]. I! has

s h o w that scheduling users one-by-one within a cell results in better performance

in terms of throughputs and power usage than simultaneous transmissions.

However, other QoS memcs like user fairness and delay deadlines warrant

explanation. Later, Zhang et a1 have reported transmission schemes with fixed

deadlines using partial channel state information [61). A finite-state Markov chain

has been used to model the channel and its state has been reported only at the end

of the t ie-s lot . Further, they have assumed that the channel transmission matrix

is k n o w well in advance.

13

Most of the reported techniques have assumed either prior knowledge on

channel state or perfect prediction of channels. l l e y have been designed with an

objective of providing either throughput maximization or user faimesddelay

bounds etc. It is also important to note that these proposals in its current form do

not exploit the time-varying property of channels. Therefore, the system

performance gains in all these approaches are highly limited.

Adaptive tmnsmissions. introduced in the late 1960's require accurate

channel estimations at the receiver and a reliable feedback path between the

receiver and transmitter [62]. Interest in these techniques were short lived, perhaps

due to hardware constraints, lack of good channel estimation techniques and/or

systems focusing on point-to-point radio links without transmitter feedback. As

these issues are less constraining in current systems and as the demand for

spectrally efficient communication is increasing, interest in adaptive modulation

methods has been revived.

Researchers have exhaustively studied the different means of adaptive

bansmissions [63-671, which include varying the transmission power, symbol

transmission rate [63], coding rate [64] and constellation size L65-671. These

schemes have provided high performance gains in terms of average specbal

efficiency by transmitting at higher throughputs under favourable channel

conditions. Sampei er a1 and Alouni el a1 have proved that further improvement

could be achieved by combining them with space diversity techniques like

multiple input multiple output (MIMO) techniques [68,69]. However, these

schemes are predetermined, fixed and may not be appropriate, if the channel

changes. or power budgetidata rates are changed. Most of these schemes : m e

used a common criterion to select the transport mode that maximizes the

throughput. Although the criteria may be suitable for best-effort services, its

performance is questionable for other guaranteed services with strict error

constraints. This demands for variable transmission powers, which play a very

active role.

Earlier, the power control problem 170-731 was identified as an eigen

value problem. The optimal power vector has been found by inversion of a

14

matrix, which was composed of users channel gains. These algorithms are

non-iterative, synchronous. and centralized. Due to the computational complexity

of these centralized power control algorithms, distributed versions of SIR

balancing techniques have been developed [7&78]. They require the mobile path

gains, which are obtained by local measurements. These approaches face two

critical problems. One of the problems is that the final SIR achieved by SIR

balancing may be unsatisfactory for a link. The other problem is that the mobiles

with different SIR requirements cannot be accommodated.

Foschini er a1 1791 have suggested a fully distributed and asynchronous

power control scheme to overcome the above said problem. Although, the SIR

requirements are met with transminer power updates, the SIR values of ongoing

calls degrades even below their required thresholds, when new admissions are

made. This in turn leads to premature dropping of ongoing calls. Moreover, if the

SIR requirements cannot be simultaneously satisfied, the transminer powers

diverge to infinity. The persistent problem with all these algorithms is the hquen t

outages at times of severe fades and thereby failed to provide any active link

protection. In order to fight against fading effects, fast power control measures

suitable for IS-95B system have emerged [80-821. However, these attempts are

partially successful in accommodating the effect of fast fadtng and thereby they

could not provide complete active link protection. Later, an effective scheme that

protects the operational links has been suggested by Wu and Bamboos et a1

[83,84]. However, the proposal lacks quicker convergence of power.

In general. power control algorithms can be classified into signal based

and SIR based power control. Signal based control adjusts transmission pc.wer

based on the received signal strength [85,86], while the SIR based power control

changes power according to the ratio of signal and co-channel interference

(pssibly plus noise) power le\,els (87-931. It has been proved that SIR based

power control yields higher performance gain than the signal based control.

However, these algorithms have implicitly assumed that calls have relatively long

holding time and they use the recent SIR measurements to adjust power

iteratively. Ulukus et a1 [94] have proposed a novel stochastic power control

dgorithm free from the above said problem. However, this algorithm is designed

15

and optimized only for constant data rate services. In summary, most of the

proposals are deterministic in the sense that they require the exact knowledge or

perfect estimates of critical quantities like S IR received interference power and

error rates. The deterministic convergence solutions are no longer valid with

random estimates of these quantities.

Goldsmith er a/ (951 have reponed considerable improvements in

throughputs with variable rate transmission techniques. Consequently, attention

has been focused on integrating various such adaptive mechanisms to effectively

support multimedia services [96-1001. Frequent outages and large power penalty

at times of serious fades are the serious limitations. Further, these schemes are not

capable of optimizing both the transmission rates and power to maximize spectral

efficiency. None of the algorithms discussed so far can effectively work in packet

switched environments. Further, the loop delay involved in the exchange of power

commands have not been considered. which leads to inaccurate results.

1.4 OBJECTIVE OF THE THESIS

An anempt has been made in the present work to study different radio

resource management tools for both TDMA and CDMA based systems to ensure

connection-quality improvement at the end user level for heterogeneous

applications The objectives of this work are as follows.

a To develop mobility aided effective call admission algorithms to minimize

the connection droppingfblocking rates besides enhancing the resosce

utilisation rates for multiple M t c classes that demands diverse QoS

requirements.

To design a reliable wireless predictor to probe the state of the time

varying wireless channel before actual transmissions, so as to ensure

effective resource usage and to minimize retransmissions.

* To introduce adaptive scheduling solutions based on the inputs provided

by the predictor to maximize system performance gain in terms of

throughputs and user fairness.

To maximize spectral efficiency by synthesizing various link adaptation

techniques such as adaptive modulation and power control.

To realize an integrated QoS approach using the suggested proposals for

improvement in performance parantees besides effective spectrum utility.

1.5 OUTLINE OF THE THESIS

The Chapter 1 deals with the evolution of various cellular systems and

standards used in wireless networks followed by a detailed discussion on quality

issues and its associated key components. The research objectives and an

extensive review of the literature relevant to the present study have also been

presented.

A discussion on various problems associated with call admission control

policies in cellular networks have been detailed in Chapter 2. A clear picture

about mobility prediction and the cellular system modei considered for the

analysis has also been outlined. The simulation study, camed out using various

algorithms in terms of the blocking rates for both TDMA and CDMA systems is

clearly presented.

An exhaustive discussion on mobile channel behaviour and channel

prediction mechanism has been covered in Chapter 3. It also includes a survey on

the proposed schedulers with and without user fairness. The simulation results are

also included.

The impact on user throughputs with the adoption of link adaptation

techmques for both access schemes has been brought out in Chapter 4. This

chapter highlights the need for interference prediction and suggests a suitable

means of estimating future interference power using Kalman filter. A detailed

presentation on various power control measures for interference limited system

has also been added. It also includes the performance analysis of the system with

joint adaptive modulation and power control schemes.

Chapter 5 mainly focuses on the need for an integrated multimedia QoS

framework that could possibly be realized by synthesizing various resource

management techniques. The suitability of the proposed QoS model has been

investigated in this chapter.

The concluding remarks of the dissertation work are brought out in

Chapter 6. It summarizes the major contributions and highlights the scope for

future work in this area.