Escolar Documentos
Profissional Documentos
Cultura Documentos
1
CONTENTS
5 Performance Evaluation 58
5.1 Performance Metrics and Simulation Parameters . . . . . . . . . . 58
5.1.1 Performance Metrics . . . . . . . . . . . . . . . . . . . . 58
5.1.2 Simulation Parameters . . . . . . . . . . . . . . . . . . . 59
5.2 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . 60
5.3 Summary of output results . . . . . . . . . . . . . . . . . . . . . 66
A Simulator validation 68
A.1 Propagation model and air interface . . . . . . . . . . . . . . . . 68
A.2 Traffic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
A.3 RRM strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
B Confidence Intervals 69
B.1 About confidence intervals . . . . . . . . . . . . . . . . . . . . . 69
B.2 Output curves with computed confidence interval . . . . . . . . . 69
2
Chapter 1
The radio resource allocation in GSM/GPRS network is the focus in this project.
Beginning with the study of GSM/GPRS network, the scope of RRM strategies is
investigated. The delimitation of the project is described in the problem shaping.
In addition, the time plan is carried out in order to organize and arrange the project
work during the semester.
1.1 Background
The wireless communication has been possible due to the electromagnetic wave
propagation through the air interface and its fast development achieves a global
communication available from one person to another at any place and any time.
The impressive growth of the cellular mobile telephony as well as the number
of the Internet users poses an exciting potential for market that combines both
innovations: cellular wireless data services. It is predicted that there will be higher
demand for wireless data services and in particular high-performance wireless
Internet access.
The overview of mobile communication system starts with several mobile radio
networks with low capacity, quality and small mobility range. These limitations
were not solved until the appearance of cellular concept used in mobile commu-
nication systems.
The main idea of cellular concept is to divide a large area into small cells in
order to reuse the frequency in the distant cell without interference. Basic on
3
CHAPTER 1. SUBJECT PRESENTATION AND PROBLEM DELIMITATION
this purpose, the first analog cellular communication system came into being.
Although the analog cellular communication system invokes the revolution of
reusing frequency for the aim of saving limited spectrum resources, a series of
problems still exist, such as only basic voice telephony, limited coverage and non-
compatibility among different networks. Instead of analog, the second generation
of digital cellular communication system has been applied. Global System for
Mobile Communications(GSM) is a typical representative among this stage and
has been very stable and widely accepted standard for mobile communication.
GSM uses circuit switched technology to transmit both voice and data.
The cellular data services do not fulfill the needs of users and providers. From
the user point of view, these data rates are too slow and the connection setup takes
too long. Moreover, the service is so expensive for most users. While from the
provider point of view, the radio resource is not utilized optimally, neither the
offer of data services.
However, the fast growth of Internet requires a wireless data access which GSM
is inefficient to support because of the fixed data service. The General Packet
Radio Service(GPRS), which is the extension of GSM, is implemented to provide
packet data service over the GSM infrastructure. The channel capacity is highly
increased and the amount of users is enlarged. The major new third generation,
called the Universal Mobile Telecommunications System(UMTS), is definitely
designed to achieve universal speech services and local multimedia systems and
is in process of development worldwide.
When considering the wireless communication system, the channel effects should
not be ignored. The performance of a wireless system is strongly affected by its
environment. In the characterization of fading channels, different components are
distinguished, such as a large-scale path loss,a medium-scale slow-varying with
lognormal distribution, and a small-scale fast varying component modelled with
a Rician or Rayleigh distribution according to the presence or absence of Line Of
4
1.2. PROBLEM SHAPING
Even though GPRS can supply many new data services as well as higher effi-
ciency in utilizing the capability of transmission network, it has to share the same
radio resources with GSM (voice) services. That means the air interface becomes
a bottleneck and the GSM/GPRS operators have to balance the quality of existing
voice service and that of newly introduced data services.
1. Set criterion for evaluating the service quality of GPRS, i.e, blocking rate
or throughput. GSM voice service quality is also in consideration though
its criterion has already been set in industry standard (blocking rate).
5
CHAPTER 1. SUBJECT PRESENTATION AND PROBLEM DELIMITATION
Since this project is designed to deal with a problem in real engineering world,
and considering the time, human resource and facility limitations the design group
has, the working scope of this project will be delimited as follows:
• Possible algorithm input and analysis model will be built on the basis of
some current existing GPRS architectures and radio resource management
strategies. That is, the developing methodology is not allowed to change the
network structure.
• Input data and traffic model will be based on the network of current GSM/GPRS
operator.
• The radio resources in this project refer to all the GSM time slots and fre-
quency bands inside one cell cluster.
In order to achieve the goal of the project, the group work has been organized
in the following phases( 1.1)
6
1.3. SOLUTION STEPS AND TIME SCHEDULE
7
CHAPTER 1. SUBJECT PRESENTATION AND PROBLEM DELIMITATION
8
Chapter 2
To introduce GSM and GPRS, first there is a need to talk about the cellular
concept on which those networks are based and then GSM and GPRS networks
will be presented to get an overview of the networks on which RRM is carried
out.
The main idea of cellular concept is the division of the area to be covered into
cells that receive a subset of the frequencies from the total set assigned to the
network. Two neighboring cells will not use the same set of frequencies to prevent
interference. The same set of frequencies can be reused at a sufficiently large
distance. When moving from one cell to another during an ongoing conversation,
an automatic channel/frequency change occurs (handover).
9
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
A group of adjacent cells which uses all of the frequencies is called a cluster.
As in Figure 2.1, 7 cells make up a cluster. All the cells labelled ”1” use the same
frequencies, but no frequency can be reused within a cluster.
10
2.1. CELLULAR INTRODUCTION
Two types of interference can cause trouble to an air born communication. Ad-
jacent interferences are caused by some other user transmitting on neighboring
channels, and guard-bands are used to fix this problem. Co-channel interferences
are caused by the use of the same channel by another customer (in a different clus-
ter). The cell shaping and power control are very important tools to help prevent
these problems. Noise can have numerous causes, but the most important ones are
the transceiver’s electronics, the background radiation, or man-made. Transmit-
ting sufficient power is necessary to have a good SNR on the receiver’s side.
Path loss means the strength of the electromagnetic wave gets weaker as it gets
further away from the transmitter. This effect not only limits the coverage of a cell
but also reduces the interferences from other cells far away. The decrease of signal
strength is proportional to the power of the distance. This power factor is usually
between 2 and 5, from free space to strong obstacle attenuation, and typically 3.5
in an urban environment.
These different types of fading will also cause temporal effects on the transmit-
ted signal, which can usually be fixed using hardware implementations or proto-
cols.
Temporal Effects
The relative delay is due to multiple echoes corresponding to all the paths the
signal can take (multipath) which all arrive with their own delays. To correct
this delay problems that could cause inter symbol interference, GSM receivers
use equalizers that are updated with a training sequence used to learn about the
11
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
channel and add up all the echoes. The use of this method makes an assumption
that the movement is slow compared to the equalizer update.
The absolute delay is a delay caused by the distance separating both ends of
the transmission. Electromagnetic waves are transmitted at the speed of light, so
the distance can have an influence if the received signals has a delay of bit times.
Two solutions are used to prevent this problem which are the use of guard bands
between each block of transmitted data, and the use of a timing advance protocol
to calculate the delay and send data at the right time.
Each of these carrier frequencies is then divided in time, using a TDMA scheme.
The fundamental unit of time in this TDMA scheme is called a Time Slot(TS) or
burst and it lasts 15/26 ms(or approximately 0.577ms). Eight TS are grouped into
a TDMA frame 120/26 (ms) (or approximately 4.615ms), which forms the basic
unit for the definition of logical channels. One physical channel is one TS per
TDMA frame.
Channels are defined by the number and position of their corresponding burst
periods. All these definitions are cyclical. Channels can be divided into dedicated
channels, which are allocated to a mobile station, and common channels, which
are used by mobile stations in idle mode.
TDMA realizes several users can share the same frequency in different time
duration.
12
2.2. GSM NETWORK
2.2.1 Architecture
The network architecture of GSM consists of three main entities called the Mo-
bile Station (MS), the Base Station Subsystem (BSS) and the Network and Switch-
ing Subsystem (NSS), each containing one or several components as shown in
Figure 2.3.
13
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
14
2.2. GSM NETWORK
The BSS, handling the radio link with the MS, has two components which are
the Base Transceiver Station (BTS) and the Base Station Controller (BSC). The
BTS comprises the radio transmission and reception devices, and also manages
the signal processing related to the air interface. The BSC manages the radio
interface, mainly through the allocation, release and handover of radio channels.
15
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
The physical layer (GSM-RF) specifies how data is transmitted from one entity
to another across the physical transport medium.
The link layer (LL) provides a link between the networking layers above it, and
the physical layer below it. It provides error detection and correction of packets
received from the physical layer. A modified LAPD (Link Access Protocol for the
ISDN D-channel) protocol, called LAPDm, is used over the Um interface. The
MTP (Message Transport Part) level 2 protocol of the SS7 protocol suite is used
over the A-interface.
16
2.2. GSM NETWORK
The similarity between the GSM protocol stack and the OSI model ends at the
link layer. Layer 3 of the GSM protocol stack does not correspond to the net-
work layer of the OSI model. GSM layer 3 is composed of 3 parts, which is
known as the message or signaling layer. It is used to set up and maintain voice
circuits between users of a mobile cellular network. It does this by managing ra-
dio resources, information about user whereabouts, and voice circuit information.
These operations are all specific to cellular radio networks because other networks
do not have to keep track of user whereabouts or movement of users from one cell
to another. The message layer in GSM is composed of three sub-layers. The Radio
Resource (RR) Management layer is in charge of establishing and maintaining a
stable uninterrupted communications path between the MSC and MS over which
signalling and user data can be conveyed. Handovers are part of the RR layers
responsibility. Most of the functions are controlled by the BSC, BTS, and MS,
though some are performed by the MSC (in particular for inter-MSC handovers.).
The RR’ layer is the part of the radio resource functionality which is managed
by the BTS. The Mobility Management (MM) layer is in charge of maintaining
the location data, in addition to the authentication and ciphering procedures. The
Communication Management (CM) layer consists of setting up calls at the users’
request. Its functions are divided in three: Call control, which manages the cir-
cuit oriented services, Supplementary services management, which allows mod-
ifications and checking of the supplementary services configuration, and Short
Message Services, which provides Point to Point short message services.
In order for these signaling protocols to be implemented over the air interface
in GSM, we will introduce the logical channels which represent the information
carried by the physical channels.
Bit Composition of the Speech Signal Recall that the RPE-LPC Encoder pro-
duces a block of 260 bits every 20 ms. It was found (though testing) that some of
the 260 bits were more important when compared to others. Below is the compo-
sition of these 260 bits.
• Class Ia:50 bits (most sensitive to bit errors)
17
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
Channel Coding As a result of some bits being more important than others,
GSM adds redundancy bits to each of the three Classes differently. The 50 Class
IA bits are encoded in a cyclic encoder (that adds 3 parity bits). The 132 Class
Ib bits (together with the 53 encoded Class IA bits and 4 trailing zero bits) are
encoded using convolutional encoding with rate 1/2 (that produce a new sequence
of 378 bits). Finally, the Class II bits are merely added to the result of the convolu-
tional encoder. The channel encoded bit sequence is now 456 bits long. Therefore,
each 20 ms burst produces 456 bits at a bit rate of 22.8 kbps. To further protect
against bit errors, the 456 bit sequence is then diagonally interleaved.
The 456 bits output by the convolutional encoder are divided into 8 blocks of
57 bits.These eight blocks are shuffled to form eight new blocks. One shuffled
block of the current speech packet and another from the previous speech packet
are written into a normal burst.
Traffic Channels
The GSM traffic channels are used to transfer information such as voice or data
at rates depending on the channel coding, and the type of traffic channel used.
The Full rate traffic channel (TCH/F) in GSM is used to convey voice or data
information in a circuit switched manner. A TCH/F is mapped on a time slot
every frame, thus allowing to transfer 114 bits of coded information every time
slot. The voice rate is about 13Kbps, as only the most important data is highly
coded, but circuit switched data information has a stronger channel coding which
allows a data rate of only 9.6Kbps. The Half rate traffic channel(TCH/H) can be
18
2.3. CIRCUIT SWITCH VERSUS PACKET SWITCH
used to transfer data information. It is mapped on a time slot every other frame,
and has half the data rate of TCH/F(4.8Kbps).
Control Channels
Control channels(Table 2.1) deal with network management messages and chan-
nel maintenance task. Any non-traffic communication between the BS and the MS
uses these channels. Three types of control channels exist which are either one or
two-way communication channels.
This table is just to give a general picture of GSM control channel but the details
inside is none of our interests in this RRM project thus it will not be discussed
further more.
19
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
periods. Circuit switch is suitable for constant bandwidth data flows, while it has
low efficiency in conveying data services with a bursting nature. Besides GSM
can only support a low data rate service like SMS.
In order to establish a comparison between GSM and GPRS, we will now focus
on the GPRS network, which uses the same physical layer and has many similar-
ities with GSM.
Unlike GSM that was designed for voice services and requires a circuit switch-
ing transmission mode, GPRS provides a packet switching transmission mode.
This feature allows an easy adoption to the bursty traffic generated by Internet
applications like e-mail, WWW and FTP. In comparison with 9.6kbps data trans-
mission rate of GSM, GPRS offers a maximum theoretical data transmission rate
of 172.4kbps. Another important goal of the technology is to make it possible
for GSM license holders to share physical resources on a dynamic, flexible basis
between packet data services and other GSM services.
20
2.4. GPRS NETWORK
2.4.1 Architecture
GPRS attempts to reuse the existing GSM platform as much as possible, but
in order to build a packet based mobile network, some network elements, inter-
faces and protocols that handle packet data are required. Figure 2.5 describes the
GSM/GPRS network which will more detailed in the following.
21
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
22
2.4. GPRS NETWORK
PCUSN The PCUSN is a stand-alone node in the BSS whose main purpose is
to complement BSCs with the specific packet processing of GPRS. It is respon-
sible for the capacity on demand feature. It decides which radio resources are
dynamically allocated to packet switched and circuit switched use. The BSC then
manages the radio resources allocated for circuit switched use, while the PCUSN
manages radio resources for the GPRS traffic itself. Its primary function is to
provide the interworking function between the radio interface (synchronous con-
nection) and the packet network Gb interface (asynchronous and connectionless).
SGSN/SIG The main functions of the SGSN are to detect GPRS MSs in its
service area, to perform mobility management, to implement authentication pro-
cedures and to send/receive data packets to/from the MS. It requests location in-
formation from the HLR through the Gr interface. These messages are routed
through the SIG, which provides the interworking between GPRS nodes in an IP
network and GSM nodes in a signalling system 7 (SS7) network.
GGSN The GGSN provides the point of interconnection with external Pub-
lic Data Networks (PDN) through the Gi interface. It stores routing informa-
tion for attached GPRS users and requests location information for mobile ter-
minated data packet from the HLR (this is accomplished transparently through
the SGSN). Its primary functions are Packet Routing and Transfer (Routing, Tun-
nelling,Encapsulation, Compression, etc.).
23
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
LLC (Logical Link Control) The LLC layer is responsible for handling the
virtual connection between the SGSN and the GPRS MS and exists even when
no physical resources are available between the two. It supports peer-to-peer data
transfer between the SGSN and the MS.
When the LLC packet arrives at the BSS (PCUSN), it is forwarded to the MS.
There the Radio Link Control (RLC) is responsible for efficient use of the physi-
cal link on the air interface and the MAC(Medium Access Control) for handling
access to the physical link.
RLC (Radio Link Control) The RLC is responsible for segmentation and re-
assembly of the LLC packets. The segmentation results in RLC blocks. Con-
trol information is added to each RLC block to allow Backward Error Correction
(BEC). The size of these segments is such that when applying the coding schemes,
they precisely fit on four normal bursts (= radio block). With BEC, both acknowl-
edge and unacknowledged mode are possible.
24
2.4. GPRS NETWORK
MAC (Medium Access Control) The MAC layer handles procedures related
to common transmission resource management. The layer allows point-to-point
transfer of signaling and user data within a cell. The medium access can be real-
ized by fixed and dynamic allocation. In MAC header, Uplink State Flag(USF) is
used as identifier.
GSM-RF This layer is based on the GSM specification, which describes the
physical characteristics of the air interface. While the RLC/MAC layer is imple-
mented in the PCUSN, the GSM-RF layer is located in the BTS.
BSSGP (BSS GPRS Protocol) This layer is responsible for the transport of
routing and QoS information between PCUSN and SGSN. It provides a connec-
tionless link with unconfirmed data transfer between BSS and SGSN. It acts as an
interface between LLC frames and the RLC/MAC blocks in the BSS, and as an
interface between the RLC/MAC derived information and the LLC frames in the
SGSN.
Among the 4 coding schemes, CS-1 has the lowest coding rate, it is also the
most robust coding scheme. Therefore, CS-1 is used for all control messages.
CS-2 and CS-3 have higher code rate and more information bits. CS-4 is the most
data efficient coding scheme and is the most vulnerable to channel impairment.
25
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
Now we can depicted the whole GPRS data unit mapping process from PDU to
time slot in Fig2.7.
26
2.4. GPRS NETWORK
27
CHAPTER 2. INTRODUCTION OF GSM AND GPRS
Control Channels
Packet Common Control Channel(PCCCH) This channel is used for a com-
mon control signalling required to initiate packet transfer. PCCCH is mapped on
one or several physical channels according to a 51 or 52 multi frame.
Four different channels are defined: PRACH (random access used by the MS to
access the network), PPCH (paging used to page a MS belonging to a given paging
group), PAGCH (Access Grant used to assign resources to a MS during the packet
transfer establishment phase) and PNCH (used to send a PTM-M notification to a
group of MSs).
28
2.5. COMPARISON BETWEEN GSM AND GPRS
• GSM
- Circuit switched transmission.
- Transmission rate per user: 13.6kbps or 9.6kbps.
- Duration-based charging.
• GPRS
- Packet switched transmission.
- Transmission rate per user: up to 172.4kbps
(Multiple time-slot can be integrated)
- Different coding schemes with GSM.
- Volume-based charging(always ”On” mode).
29
Chapter 3
In this chapter we present the play scenarios of existing RRM strategies first,
and then describe in more detail the Best Effort policy and our proposed RRM
strategy, to investigate how we can improve the GSM/GPRS system performance
comparing to Best Effort.
3.1 Introduction
Radio Resource Management is the function responsible for the establishment,
the maintaining, and the release of voice calls (CS) or data transmission (PS). The
establishment of connections is about allocating resources or blocking the incom-
ing CS or PS calls if necessary. Maintaining a connection requires dealing with
handovers and power control, but also reassignment of resources and transmission
scheduling for GPRS.
30
3.1. INTRODUCTION
available, resources can be taken from existing GPRS transmissions. This process
is called preemption.
Uplink: Two different strategies can be used to assign resources in uplink. The
Fixed Resource Allocation (FRA) strategy will send the TBF to the mobile at the
establishment of a connection with directions on when it is able to send blocks
on the given time slots. The Dynamic Resource Allocation (DRA) strategy will
send the TBF information to the mobile with an identifier called Uplink State Flag
(USF). The mobile will then listen to its allocated downlink TS and checks there
for its USF to know when it is allowed to send its uplink blocks.
31
CHAPTER 3. RADIO RESOURCE MANAGEMENT
For GPRS users requiring K channels (depending only on the capacity of the
mobile as QoS is not considered), a TRX will be chosen which will be the last
least-loaded one in terms of number of TBF and GSM calls assigned. In this
TRX, the K least loaded time slots will be selected and if less than K TS are
available, the maximum amount possible will be allocated. The selection of Time
Slots to be allocated for the TBF is illustrated in Figure 3.1.
32
3.2. ”BEST EFFORT” STRATEGY
33
CHAPTER 3. RADIO RESOURCE MANAGEMENT
When resources become free after a congestion period, the priority of TBF
assignment goes first to the TBFs having the least resources to improve the user
throughput and release resources more rapidly and then to the access queue to
decrease access delay and GPRS blocking rate.
3.2.3 Scheduling
Once a GPRS call has been allocated resources, a scheduling technique is used
to choose blocks that has to be sent in TS holding multiple TBF (multiplexed time
slots). In the ”best effort” strategy, in order to send blocks fairly in multiplexed
time slots, the queue used is a Round Robin queue as illustrated in Figure 3.3,
meaning that blocks will be sent from a different TBF every block period.
34
3.3. PROPOSED OPTIMIZED RRM STRATEGY
In the simulated scenario, the grading is based on the Carrier to Interference ra-
tio (C/I). The main purpose to choose C/I is to try to improve the system through-
put. Since the link adaptation is employed, the throughput is always considered
optimized for a chosen CS-x, which is elected according to C/I. In the case we
give priority to high C/I values, the throughput per cell should increase and re-
sources should be released more rapidly, thus decreasing the access delay and the
blocking rate.
The grading could be based on other parameters such as QoS. The QoS prece-
dence and required access delay could be used for grading a request in the access
queue, thus giving access priority corresponding to the requirements. Moreover,
35
CHAPTER 3. RADIO RESOURCE MANAGEMENT
the QoS mean and peak throughput classes (in addition of precedence and C/I)
could be used to compute grades for resource allocation and scheduling.
Access Queue If no free TS is available for GPRS data, the incoming request
will be put into the access queue. The principle for the resource allocation is
FIFO with priority, where the priority is based on the grading(G) of GPRS data
mentioned before. If the grades of the requests are different in the access queue,
the highest one will get the available TS first. When several requests have the
same grade, FIFO strategy will be applied. Moreover, when the access queue is
filled, the request with smallest grade is blocked instead of blocking the incoming
request.
Below a significant example is explained. The next request waiting in the queue
can be allocated in one of the two TRXs that are occupied as in the figure:
36
3.3. PROPOSED OPTIMIZED RRM STRATEGY
Each TS has a grade(T SG ) calculated as the sum of all the TBF grades assigned
to this TS. The possible % of a TS utilization for an incoming TBF is calculated
by the following expression:
GincomingT BF
T Savailable = P (3.1)
T SG + GincomingT BF
In this example, the grade of the first request waiting in the access queue is
12. For the first TRX, there are 6 TSs allocated for GSM, and only one is free.
The TS utilization of this TRX by this TBF would be 1TS. In the second TRX,
the TS utilization would be 0.93TS. The calculation of the T SG is shown in the
picture. Although the second TRX has more TSs available to be used, the actual
TS utilization would only be 0.93TS, which is less than for the first TRX. In this
case, the first TRX is selected.
3.3.2 Scheduling
In the case of multiple TBFs occupying one TS, as mentioned before, the
scheduling of the optimized RRM strategy is also based on the grading which,
in our simulation, depends on the C/I in the propagation channel. The probability
for each TBF in the same TS to use the assigned resource is calculated from its
37
CHAPTER 3. RADIO RESOURCE MANAGEMENT
For example, there are 3 TBFs in one TS. The grades are 3,1 and 6 respectively.
Therefore, the possibility for the three TBFs to obtain the TS is 30%, 10% and
60%, which guarantees the highest graded TBF will have most time to get TS to
transmit data.
In the next chapter the simulation platform used to test and compare the two
explained RRM strategies is described. The output results will be analyzed and
compared to validate the proposed optimized algorithm.
38
Chapter 4
In this section, the network simulator and simulation assumptions are presented.
The simulator is a time-driven network level simulator based on events,where the
basic simulation time step is 20ms (1 RLC block time of GPRS), but if events
come between two time steps the simulation will jump at that time before going
to the next block (event-based simulation). Only the downlink traffic of 1 cell is
investigated and simulated. Hence the simulation scenario was limited to Base-
Station and Mobile-Stations and up till RLC layer.
Since the purpose of this RRM algorithm is to improve GPRS performance and
only take GSM voice call parameters as its limitation, the role of GSM voice call
was limited to occupy some radio resources. In this case, blocking rate was the
only service quality identifier of GSM voice and call dropping and handover was
not in our consideration. Therefore the transmission time step was fixed to 20ms
(1 RLC block time of GPRS, made up of 4 TDMA frames).
Figure4.1 shows the channel with the radio resources available in a medium
capacity cell without applying any partition for dedicated channels for voice or
data.
39
CHAPTER 4. SIMULATION MODEL OF RRM
According to the mobility, a Typical Urban scenario with mobile user speed
of 3Km/h (TU3)[?] was assumed, since it is a typical case for pedestrian user.
The simulated mobiles are uniformly distributed in the cell and their movement is
reduced to a small area around their initial position when setting up a voice call
or data transmission session, so no mobility tracking was simulated.
All the mobiles are capable of using at most 4 TSs in one frame. Also all the
mobiles are in the active state to receive a GPRS data transmission.
The parameters from general cellular assumptions are presented in Table 4.1.
40
4.2. RADIO CHANNEL MODEL
Due to the complexity of a cellular system, in most case link level and net-
work level simulators make up of a entire simulator but are studied and operated
separately. A link level simulator generate a point-to-point MS-BSS transmission
process and the result (BER, BLER, etc) will be included in the network level
simulator. Since the mapping curves from C/I to BLER in given coding schemes
and propagation models have already been validated in the link level simulation in
[?], we propose a C/I generation model that includes the mapping process in our
network level simulator. An overall picture of our radio channel model is depicted
in Figure 4.2.
41
CHAPTER 4. SIMULATION MODEL OF RRM
Path Loss
In free space, the path loss is proportional to the distance square [?]. In an urban
area, it is typically proportional to r−3.5 . The detailed locations and movements
of each mobile were not modelled, instead each user was assigned a random ini-
tial distance from BTS according to their spatial distribution function. Assume
MS were evenly distributed in the hexagonal cell, we got the distance Probability
Density Function(PDF):
where r is the distance between MS and BTS and k is a coefficient after normal-
ization.
Shadowing
Shadowing or slow fading can be seen as the gradual changes of the local mean
received power for a MS, where the local mean power represent the average over
a few tens of wavelengths [?]. The shadowing effect can be described by a log-
42
4.2. RADIO CHANNEL MODEL
In the situation of TU3, where the speed of the mobile is slow (3km/h), the
effect of shadowing causes the signal attenuation nearly flat during approximate
1000 RLC blocks, where the wavelength equals to 30cm and 1 RLC block lasts
20ms. In this case the fading level of shadowing was fixed to one value during
one session in the proposed simulator.
Multi-path fading
In corresponding to the slow fading, the multi-path fading is defined as fast
fading. It can be described as a Rician or Rayleigh distribution [?]. In the former
case, there is a significant Line of Sight(LOS) between BS and MN; while if no
LOS exists, it is referred to the latter case. The Probability Density Function(PDF)
of Rayleigh distribution is expressed as:
r r2
f (r) = exp(− ) (4.3)
σ2 2σ 2
The fast fading characterizes the fast variation of the signal envelop over a short
distance of a few wavelengths or over short time durations of seconds [?]. In
TU3 situation, the variation of the signal changes relatively slow, nearly 50 RLC
blocks unchanged . Detailed analysis on the time correlation and frequency cor-
relation of multipath fading is presented in Appendix A. It was proposed that
given a frequency hopping distance greater than 500kHz every TDMA frame [Ref
Marco](separations of 600to 800 KHz are reported to be adequate [?]), the mul-
tipath fading levels for consecutive RLC blocks will be fully uncorrelated, that
means only random block error need to be simulated.
In reference [?], fast fading effect has been included in the proposed CIR-BLER
mapping tables and mean BLERs during one session were given. Hereby only
pathloss and shadowing need to be implemented in proposed simulator.
Co-channel Interference
We consider Co-channel interference as the predominant aspect of C/I, which
includes the 6 interferes from the first tier(6 adjacent cells with the same fre-
quency), and neglects all other interferers. This model is depicted in Fig 4.3.
43
CHAPTER 4. SIMULATION MODEL OF RRM
Assume the same transmitted power for both desired and interference signal,
based on pathloss, the CIR is
1
CIR =
2(Ru − 1)−3.5
+ (Ru − 0.5)−3.5
+ (Ru + 0.5)−3.5 + (Ru + 1)−3.5 + Ru−3.5
(4.4)
[?] Where Ru=D/r, D: reuse distance, r: MS-BTS distance.
Conclusion
The situation of propagation simulation is under TU3 model, where the fast
fading plays a dominant role in the effect of propagation channel, given the slow
speed of the mobile. Since the values of BLER are taken from the mapping ta-
ble in [?], where the multi-path fading has already been applied, and shadowing
level during one session will keep almost constant, the path loss effect will play a
predominant role in deciding the C/I value for each session.
When considering different users with the same distance from BTS, with the
same path-loss effects, log-normal fading can be considered as a spatial variation
among different users with the same distance from the BTS. Assumed the received
desired signal and the interferers (which have experienced path-loss) face to in-
dependent log-normal distributed stochastic variations, which can be presented in
dB as Equation 4.5.
X
C/I = DesiredSignal − Interf eres (4.5)
44
4.2. RADIO CHANNEL MODEL
and both of the components on right side follow normal distribution, C/I follows
a normal distribution N (µ, δ)(since the sum of independent normal distribution is
still a normal distribution), so the C/I is also a log-normal distribution. The mean
value was calculated from path-loss and the standard deviation was set to 7dB.
Finally, the whole propagation model can be presented as the diagram in Figure
4.4, in which the average BLER of a user can be calculated.
When we link the user distribution and the corresponding CIR mean value from
path-loss(before imposing log-normal fading), the PDF of mean CIR in a cell can
be shown in Fig 4.5.
45
CHAPTER 4. SIMULATION MODEL OF RRM
After C/I for each session was generated, the next step is to map it to BLER
according to the mapping table[?] under a selected coding scheme. The process
46
4.2. RADIO CHANNEL MODEL
to select the most appropriate coding scheme is referred to as link adaption (LA).
We have assumed an ideal case where we can measure or estimate the C/I of
each user before the real data transmission start, so we are able to select a coding
scheme to achieve the maximum session throughput based on figure 4.6 [?] .
47
CHAPTER 4. SIMULATION MODEL OF RRM
The Traffic Models are defined to characterize the behavior of the users in a
cellular system. So the time variation of traffic is studied, both for voice and
GPRS data. Although GPRS has a wide range of applications and its usage is
increasing, currently the presence of GSM voice service is much higher than data,
so their representative probabilities are assumed over the global traffic load, e.g.
the 70% voice traffic and 30% data load [?]. This values can be varied later on in
the simulation to evaluate the results with different traffic load probabilities.
The term of session is referred to both voice and data transmission. The session
is defined from the moment when an event arrives to the BSS until the transmis-
sion is finished, such as for voice the end time is reached or for GPRS all the
RLC blocks have been sent to a specific Mobile Node or in the worse case when
this event has been blocked or dropped. For GPRS data, a session consists of a
number of information bits,corresponding to 1 TBF, simplifying the hierarchical
layers for GPRS data.
Moreover each generation session is associated with one MN using uniform dis-
tribution. One MN representing only a propagation condition at a specific distance
from the BS, it is considered to be able to accept as many sessions as possible. It
can correspond to the situation with some real mobiles being at the same distance
but with different shadowing factor, which is calculated at the beginning of the
session.
48
4.3. TRAFFIC MODEL
• arrival time. The number of events arriving in one radio cell is described
as a Poisson distribution and it requires the definition of the mean arrival
rate,λGSM . It models the number of random occurrences of voice calls
(N(t))in a specified unit of time. The probability that the variable N(t)
reaches a value n is given by the formula:
49
CHAPTER 4. SIMULATION MODEL OF RRM
However, this aproach to the higher level is not necessary since the session
is assumed to be composed by a number of bits whose size can be calculated
according to a certain distribution function.
As it was seen for GSM calls, a traffic model consists of two parts: the arrival
process and the activity phase.
• The arrival process is described most frequently with the Poisson function
which counts the number of events coming within a certain time period.
The mean arrival rate per session,λGP RS is specified and assigned a value.
As it was defined for GSM calls, an exponential distribution for the inter-
arrival time process is used to determine the exact time when an GPRS event
arrives within the simulation time.
• For the activity phase, the duration of the session is characterized with
the number of bits transmitted and calculated by means of the geometric
process specified below. Then, according to the number of bits obtained
and the coding scheme specified, the number of RLC blocks is carried out
(Fig.4.8). The testing of the following distribution function will be carried
out in Chapter 5.
The Geometric distribution is characterized as follows (parameter 0<p<1)[?]:
The probability density function:
The variable x is referred to the number of bits, and this distribution is used
first because it is simpler to model.
50
4.3. TRAFFIC MODEL
Under those traffic conditions, the simulator is implemented, and the values
of these parameters will be specified.This traffic generation for voice and data is
shown in Fig.4.9.
51
CHAPTER 4. SIMULATION MODEL OF RRM
(λGSM ) and the average holding time (1/µGSM ), where Aof f −GSM is called the
offered traffic for voice.
1
Aof f −GSM = λGSM ∗ (4.9)
µGSM
For GPRS, the offered load (Aof f −GP RS ) is calculated with the arrival rate
(λGP RS ), and the mean holding time (1/µGP RS ), which is estimated by means
of average values of the session size and the data rate in coding scheme . The
session size (number of bits per session) is obtained with the previous distribution
function and together with the coding scheme selected in the propagation model,
the number of RLC blocks is determined. The mean holding time is calculated in
the following equation:
meansessionsize
1/µGP RS = (4.10)
meandatarate
The overall offered load is simply carried out by adding the voice offered load
to the data offered load [?] :
This mapping from the traffic load to the traffic parameters (λGSM , 1/µGSM ,
λGP RS , session size) is shown as follows in Fig.4.10:
52
4.4. RLC/MAC FUNCTIONALITIES
Therefore, in the simulation the input parameters the user can introduce linked
to the traffic model are the traffic load for both voice and data, the mean holding
time for call voice, and the probability for the distribution function in the session
size (See Table 4.4):
4.4.2 ARQ
In order to ensure reliable, verified, error-free delivery of data in the presence
of different propagation conditions, the simulated RLC layer is working in the
acknowledged mode so that automatic repeat request (ARQ) schemes can be used
to synchronize and acknowledge the transmission of data between the base station
and the mobile node.
53
CHAPTER 4. SIMULATION MODEL OF RRM
This ARQ protocol insures a good use of the channel capacity as only erroneous
blocks are retransmitted.
54
4.5. SIMULATOR STRUCTURE AND BASIC DESCRIPTION
The traffic model object generates all traffic before the simulation start accord-
ing to the earlier described model. It links a generated session with a mobile in a
uniformly distributed manner and communicates the events to the BSS object at
the right time.
The air interface object contains the actual transmission process and links the
scheduler of the BSS with the different MS to compute how blocks are transmitted
and return the different acknowledgment to the BSS (simplified uplink part).
The BSS object contains the RRM (channel allocation, scheduling, and channel
release process), and the handling of acknowledgments. It’s divided into 3 parts,
an access part, handling the arrival of a session from the traffic model object, a re-
source update part dealing with updating active session queues and (re)assignment
of resources when necessary, and a scheduling part, linked with the air interface
object and preparing the next transmission.
The trace object is linked with most of the other objects and keeps track of the
sessions to be able to output the results at the end of the simulation in a matlab
compatible file. Outputs are then plotted with some matlab functions using the
data traced from this object.
55
CHAPTER 4. SIMULATION MODEL OF RRM
56
4.5. SIMULATOR STRUCTURE AND BASIC DESCRIPTION
Initialization Before starting the simulation, all the MS objects are initialized
(propagation condition), and traffic for all the simulation is generated according
to the specified traffic load parameters.
Next block time step When a block time is reached, first the transmission of
all blocks is simulated using the air interface object. The resource update part
of the BSS is then called with earlier acknowledgments to update the active ses-
sion queues and release if a session or GSM call has ended, assign or reassign
resources if necessary. Finally, the scheduler of the BSS is called to compute the
next transmission.
57
Chapter 5
Performance Evaluation
This chapter mainly aims at analyzing the simulation result of the proposed
RRM strategy. Validation of the simulator can be found in Annexe 1. First of all,
the choice of the performance metrics used to evaluate the proposed strategy in
comparison with the Best Effort are described, as well as the simulations parame-
ters. Then the simulation results are analyzed for each considered metric in order
to validate the expected improvements of this strategy.
58
5.1. PERFORMANCE METRICS AND SIMULATION PARAMETERS
59
CHAPTER 5. PERFORMANCE EVALUATION
Confidence Intervals In order to analyze the validity of our results, 95% confi-
dence intervals of the shown output figures have been observed in Annexe 2.
Throughput Per Cell Figure 5.1 shows the throughput per cell of both strate-
gies.
60
5.2. PERFORMANCE EVALUATION
61
CHAPTER 5. PERFORMANCE EVALUATION
It can be observed that up to a certain traffic load, the throughput per cell in-
creases with the traffic load. Until higher traffic loads, the system throughput is
equivalent for both strategies. That can be explained by the fact that for low traffic
loads, there are not many GPRS users in the cell, whereas this number increases
with the traffic load.
The enhancement provided by the proposed RRM appears as traffic load in-
creases. It can be observed that the maximum throughput is higher for the pro-
posed RRM than for the Best Effort. This performance enhancement is due to
the fact that our grading system prioritizes the GPRS users with higher CIR in
the scheduler and the access queue, thus users that are using Coding Scheme with
higher throughput.
After a certain point is reached, the throughput decreases as the number of GSM
calls increases and voice has the priority. We can notice that this point appears for
a higher traffic load on the proposed strategy.
Mean throughput per user Figure 5.2 shows the mean throughput per user of
both strategies.
62
5.2. PERFORMANCE EVALUATION
63
CHAPTER 5. PERFORMANCE EVALUATION
The mean throughput per user decreases with the traffic load as fewer resources
are available. For low traffic load, the result is equivalent for both strategies but it
can be seen that the trend is sharper for the Best Effort. The mean throughput per
user is thus better with the proposed strategy for high traffic loads. The reason for
that result is the same as for the throughput per cell.
Mean access delay Figure 5.3 shows the mean access delay of both strategies.
The access delay for both strategies is increasing with the traffic. As the more
users there are, the more filled is the access queue. The performance result is much
better for the proposed algorithm because it prioritizes the better CIR sessions thus
the average data transmission time is less and resource are released faster allowing
other sessions to get out of the access queue.
GPRS Blocking Rate Figure 5.4 shows the GPRS Blocking Rate of both strate-
gies.
64
5.2. PERFORMANCE EVALUATION
Figure 5.4: Best Effort: GPRS Blocking Rate Vs. Traffic Load
The GPRS blocking rate is much higher for best effort for the same reasons as
for the delay. Indeed blocking a session happens as the access queue gets filled.
In order to check this hypothesis, the throughput according to CIR was output
in Figure 5.5. It is a mean for all simulated traffic loads.
65
CHAPTER 5. PERFORMANCE EVALUATION
This figure shows that the hypothesis that users with bad CIR have a less
throughput with our strategy than with the best effort is wrong. In average, every
GPRS user should be more satisfied with this algorithm.
66
Chapter 6
This project mainly aimed at investigating an optimized RRM strategy for GSM/GPRS
like networks in order to improve the system performance. Based on this purpose,
the principle of GSM and GPRS network has been studied first. A simulation
platform has then been implemented with basic propagation and traffic model as-
sumptions. The currently used RRM strategy (Best Effort) has been implemented
as well as a newly designed optimized RRM strategy. Both strategies have then
been compared and analyzed to conclude that RRM schemes are a very important
feature of wireless networks and could be enhanced in the future especially for
UMTS like networks.
Due to the time limitation, the RRM strategy developed is very simple and
could be improved in future work by considering a more optimized grading crite-
ria that could take into account QoS for example. In addition to that, the effect of
dedicating GPRS or GSM channels would also have to be investigated.
67
Appendix A
Simulator validation
68
Appendix B
Confidence Intervals
The confidence level is the percentage of the intervals (computed for each in-
dependent sample) that are likely to include the searched mean value. The confi-
dence level considered in the following section is 95% (most used value).
69