A Link-to-System Level Interface for B3G Scenarios

Y. Nasser1, M. Helard1 V. Monteiro2, J.Bastos2, J. Rodriguez2, H. El-Mokdad3

1: Institute of Electronics and Telecommunications of Rennes, UMR CNRS 6164, Rennes, France.

2: Instituto de Telecommunicaçoes, University of Aveiro, Campus de Santiago, 3810-094 Aveiro, Portugal

3: Lebanese University, Hadath Campus, Beirut, lebanon

Email: youssef.nasser@insa-rennes.fr; vmonteiro@av.it.pt

ABSTRACT— Advanced simulation tools have taken a lot of attention for development and analysis of new and existing protocols and technologies in wireless communications. The huge amount of information, protocols and techniques to be analysed make however their implementation difficult in one simulator. It is thus of paramount importance to define a common interface between the different layers in order to simplify the implementation task but guaranteeing also the confidence of the obtained results. This paper provides a complete specification of the system level simulation envisaged for Beyond 3rd Generation (B3G) systems. The main idea of this paper is to explain how we could extract information from the link level simulation and to implement them in the system level simulation. Two particular cases are considered for the proof of concept in this work. The first one concerns the Spatial Division Multiple Access (SDMA) technique using positioning information. The second one concerns the use of the cooperative communications to improve the quality of service at the cell border.

Index Terms- System-to-link level, Cooperation, Resource allocation, MAC protocol, Cross-layer.

1. INTRODUCTION Efficient use of radio resources requires Cooperative Radio Resource Management (CRRM), a module that carries out RRM on a global scale between systems of diverse technologies and operators. To solve the CRRM challenge, an experimental platform is required that models all environmental and system issues pertaining to a heterogeneous networking scenario, and that has desirable attributes which include: low complexity and simulation time and high modeling accuracy.

A traditional system level simulator simulates a large number of mobile terminals in a wide environment with several base stations. In practice, the use of an efficient experimental platform dealing with all technologies, protocols, simulation scenarios and all mobile terminals is rather time consuming. Nevertheless, it is important to include the different transmission parameters (mobility, fading, interference, etc) in the system level simulator. In literature, some work has been done to implement and analyze the system performance with all these scenarios and parameters, in terms of the average of the Signal to Interference and Noise Ratio (SINR). The problem is that the average SINR does not fully reflect the Quality of

Service (QoS) at the user terminal. It is therefore essential to express the QoS requirements in terms of Block Error Rate (BLER) (or Packet Error Rate, PER) which reflects better the actual quality of the signal received by the user terminal.

This paper presents an experimental link-to-system level interface for B3G scenarios. The goal is to describe first the system level platform used for RRM algorithms in B3G scenarios, using wireless systems. Afterwards, we describe one of the most promising techniques used for link-to-system level interface called effective exponential SNR mapping (EESM) technique. This technique, proposed in 3GPP-LTE, is applied in this work in different transmission scenarios for the proof of concept targets. The latter will be considered in two cases: in SDMA technique and in cooperative communications networks.

The rest of the paper is presented as follows. Section 2 presents the system level simulator architecture based on a layered structure of communications systems. Section 3 describes the EESM technique as the link-to-system level interface. In section 4, we show the proof-of-concept by considering two transmission cases: SDMA technique and cooperative communications technique. The conclusions are drawn in section 5.

2. SYSTEM LEVEL SIMULATOR ARCHITECTURE

2.1. Problem domain

The high level objectives of the system level tool are to measure system coverage capacity and spectral efficiency, for which the evaluation criteria is given in [1][2][3]. Moreover, the design of the reference system level simulator must be sufficiently complete, so as to provide sufficient modeling accuracy, whilst still keeping simulation time and excess complexity to a minimum.

In order to reflect a realistic system, the performance evaluation should consider the impact of the relevant layers of the communication protocol: physical layer, link-layer (L2 layer) and upper layers. The details regarding the Link Level Interface to the Physical layer can be found in [4]. The structure of a single Radio Access Technology (RAT) simulator, with some of the blocks representing the functions described above, is presented in Figure 1.

The Medium Access Control (MAC) layer comprises two types of models: MAC protocols that include algorithms and procedures which affect system performance and optimization, such as Call Admission Control (CAC),

978-1-4244-7157-7/10/$26.00 ©2010 IEEE 233

Handover, Dynamic Channel Allocation (DCA); and another group related to the modeling of the system in order to validate the MAC protocol/algorithms, such as mobility models, service models and traffic queues, radio channel propagation models and the actual area space considered in the simulations.

Figure 1- System-level Simulator structure for the OFDMA-MIMO.

More specifically, the MAC components include:

2.2. MAC protocols

Packet Scheduling. The scheduler decides how to allocate the appropriate radio resources to each user based on the following context information: service type, user QoS profile, and channel performance. In WCDMA, four types of scheduling are defined:

• Time division scheduling: This is based on the concept of several users sharing the same transport channel in the time domain. Thus each user will be allocated the entire bandwidth for a short period of time, each sharing the same code. This technique provides code efficiency, and is more suitable for bursty traffic. In addition, it can provide appropriate link performance due to the high data rate. This scheme is usually used with shared channels. This type of allocation will provide high interference variations with time, thus having impact on real time services.

• Code division scheduling: Each user is given bandwidth on demand, by allocating the users with different codes. The scheme is associated with dedicated channels, and low bit rate users. It will provide an initial delay on set-up, and can lead to more predictable interference loading. The efficiency of this type of scheduling is dependent on the accurate estimate of the average bit rate. A poor estimate will lead to inefficient use of the spectrum, thus a dynamic allocation scheme is desirable.

• Power based scheduling: It will allocate resources based on the user location, assigning low bit rates to users near the cell edge, and higher bit rates to those nearer the

base station. This scheme will have direct improvement on the average downlink capacity.

• Sub-channel based scheduling: This is a specific scheme for OFDM systems. Allocation is performed on the basis of sub-channels. A number of sub-channels is allocated per used as a function of the fading affecting these bands. The achieved improvement is in terms of bandwidth required by the user application while optimizing band utilization.

Although all the schemes can provide performance improvements in different conditions, there is no single scheme that can be considered to be the best candidate. Typically, a combination of scheduling techniques isused to provide overall performance gain. In this paper, the scheduler is based on time division although it can be extended to consider allocating resources both in time and frequency. Still, only dedicated transport channels will be considered in the reference stage, and channel signaling time set-up will be implicitly assumed, but will not be considered in the overall delay associated with dropping a packet session.

In packet based systems, scheduling refers both to selecting packets based on priorities primitives, and mapping them into resources (time slots, coding and carriers), using cross-layer information whose content is delay requirements for the service (from upper layer) and suitable slot/carrier for that service.

Automatic Repeat Request. Simple Automatic Repeat Request (ARQ) is considered for non-real time services. It is assumed that variable IP packet sizes are translated to fixed packet sizes in the Radio Link Control (RLC) layer, through segmentation, concatenation and padding. When the link quality is below the target level, the QoS block will decide whether to drop any packets based on the average SIR value measurement and target value. Packets that are assumed to arrive with error will be dropped and retransmitted. The retransmission is implicitly assumed, and the delay counter associated with the user queue will be incremented accordingly. We assume that many packets can arrive within the time interval. If the SIR is below the target value, a PER model will suggest whether the specific packet is in error.

2.3. System models

Mobility models. Typical models are being employed to model mobile movement in indoor, outdoor urban, and sub-urban environments. Parameters associated with mobility include speed, probability to change speed at position update, probability to change direction, and the de-correlation length. The latter will dictate the simulator time interval between mobility updates. A detailed specification is given in [1].

A simulator map provides a description of the cellular map, which includes the cell descriptions, base station locations, and the manner in which it will model mobile movement at the system boundaries. A wrap around model is being used instead of modeling mobile movement bouncing of the edges of the outer-cells. This means that the mobile may migrate off the edge of the system boundary and, emerge on the opposite side, in a wrap around fashion.

234

QoS measure. This module is responsible for analyzing the link quality for each transport channel. If the quality deteriorates below a certain level, then it will take the appropriate action. It will increment the service delay counter, and will drop the packet session if the maximum delay has been exceeded. The detailed definition of the dropping criteria is given in [1][2][3].

Service Queue. All services are packet based, and defined by the QoS context, that will include information such as instantaneous bit rate, average bit rate and current delay, and maximum tolerated delay. All new incoming users will be assigned a priority value, and then placed in the queue. This service queue will list all the mobiles that are waiting to be served, as well as all users that have already been allocated a transport channel. The QoS Control block will look at this table to check whether any user has breached the QoS, and drop it from the system.

Dynamic Channel Allocation (DCA). The DCA algorithm is considered since it provides extra performance tracking the channel variations. It is important to validate the basic simulator architecture at the earliest design stage, and to provide some benchmark performance curves. In this way, the immediate improvement given by DCA can be noticed at the intermediate design stage, and verified. The need for DCA arises when changes either in the traffic, or channel conditions lead to under occupancy and a reduction in the QoS.

Propagation Module. The module will model path loss and slow fading. Channel models for indoor environments, outdoor urban and rural environments will be provided.

Link Level Interface. To provide an adaptive solution, the system level platform must be integrated into the Link Level platform. This solution is not efficient, and there is a direct trade-off between modeling accuracy, complexity and simulation time. Therefore, the PHY layer is typically modeled by a Link Level Interface in the form of look-up tables, which models the average link performance for a given scenario defined by the channel, interference models, mobility and service. Moreover, an interface translates the system level parameters into the appropriate transport format parameters to simulate the Link Level chain, resulting in a table with SIR vs. PER (Packet Error Rate) for a specific simulation environment.

Mobiles. The system will have the flexibility to support different mobile types, supported by the inheritance attribute that object oriented programming offers. Each mobile type will be defined by the following parameters:

• Antenna type: antenna type will be assumed to be omnidirectional;

• Maximum transit power: the maximum transmit power the mobile can support;

• Mobile noise figure: the receiver sensitivity; • Power dynamic: the transmit power range the mobile

can support between a maximum and a minimum; • Mobile coordinates: each mobile is responsible for

updating its coordinates, in terms of position and velocity.

In the reference stage, it is assumed that the same mobile type is considered for all the scenarios.

Base Station/APs. As in the mobiles case, the Base Station class is a template, which will support child objects with added functionality. This generic template can be defined as:

• Antenna type: 4 antennas for a tri-sector divided area; • Maximum transmit power: the maximum transmit

power the mobile can support; • Base Station noise figure: the receiver sensitivity; • Power dynamic: the transmit power range the base

station can support between maximum and minimum;

• Resource Unit Identifier: A three dimensional coordinate provides a description of the frequency slot, time slot, and code number.

Signaling: All signaling is implicitly modeled to reduce simulator processing overhead.

Transport Channels. Transport channels reflect the available resources in the cell. Separate resources exist for both uplink and downlink. The capacity of the resource unit is dependent on the receiver and frame structure, as well as on channel link quality.

3. LINK-TO-SYSTEM LEVEL INTERFACE The main problem in the link-to-system level interface is how to model and include all transmission conditions and parameters in one physical link state information (LSI). In literature, different estimations algorithms were proposed for OFDM systems. The Quasi Static Method (QSM) [5] was the first approach proposed to evaluate the link level performance. It is based on the computation of the average SNR value obtained at the output of the Fast Fourier Transform (FFT). However, the average SNR could not be suited for as a LSI in real scenarios. Indeed, it does not take into account the channel coding. Moreover, the specific channel realization may result in a performance which is different from that predicted by the average SNR technique. If we consider two average SNR values and such that

, the average SNR approach does not guarantee that the estimated PERs at the output of the channel decoder satisfy PER1< PER2.

Another solution to predict accurately the PER at the output of the channel decoder is the use of the EESM [6][7][8]. The results given in different systems [7][9][10] verify the proposed PER estimation technique at the output of the channel decoder.

The EESM technique is deduced from the Chernoff union bound. Let N denote the packet size in complex data symbols. In general, the data symbols in the packet are transmitted over different resource elements (e.g. sub-carriers), and therefore they may experience different propagation and interference conditions. Thus, the data symbols may have different SNR values. Let SNR be the vector of N instantaneous SNRs received at the output of the detector. The problem of determining an accurate BER prediction method comes back to looking for a relationship such that:

235

( 1)

where Pe denotes the bit error probability (BEP) and f is the prediction function, which should be invariant with respect to the fading realization, to the multi-path channel model and should be applicable to different MCS in a soft way, i.e. by changing the values of some generic parameters [6]. In the context of an AWGN channel, a direct relationship ξ exists between the SNR and the error probability.

, (2)

The function ξ is called the mapping function. It is obtained through theoretical analysis or system level simulation with an AWGN channel. In the general context of a fading channel, where the SNR varies, the function f in ( 1) can be written exactly as a compound function of the AWGN function and a compression function r [7]:

with (3)

The function r is referred to be as the compression function since its role is to compress the vector SNR of N

components into one scalar SNReff ( . The scalar SNReff is called the effective SNR and it is defined as the SNR which would yield the same error probability in an equivalent AWGN channel as the associated vector SNR in a fading channel. By writing (3), we have merely turned the problem of determining the evaluation function f into the problem of determining the compression function r.

In an OFDM system, it was concluded that the key issue to accurately determine the appropriate PER after channel decoding is to use the effective SNR in combination with AWGN curves. In [7], the EESM technique is proposed, which is based on the Chernoff Union bound [6], to find the effective SNR. The key EESM technique expression relevant to an OFDM system is given by:

log 1 e (4)

where | |²² is the SNR received at the output of the

nth sub-carrier of the detector which must be estimated from the system level simulations and λ is a unique parameter which must be estimated from the system level simulations for each modulation and coding scheme (MCS). It is estimated once by preliminary simulation for each MCS. When the SNReff is computed, it will be used for PER prediction at the output of the channel decoder with a simple look-up table (LUT), as shown in Figure 2. This LUT gives the PER at the output of the channel decoder as a function of the SNR for a Gaussian channel. It is computed analytically or by simulations. The uniqueness of λ for each MCS is derived from the fact that the effective SNR must fulfill the approximate relation:

(5)

where is the Packet Error Probability (PEP) for the AWGN channel which depends only on the MCS.

Figure 2- PER prediction through EESM

Using the effective SNR of (4), we are now able to evaluate the BER using the LUT as shown in Figure 2.

4. PROOF OF CONCEPT

The objective of this section is to judiciously prove the concept of the simulator system architecture as well as the link-to-system level interface through two real scenarios.

4.1. First Scenario: Using SDMA Technique

SDMA exploits MIMO transmit techniques to increase cell capacity by facilitating the allocation of users in the spatial domain. Beamforming is based on the principle that an antenna pattern is steered by applying a weight, i.e. a complex value to each antenna element. The pattern weight is represented by a weight vector, which contains one weight per antenna element. The linear nature and number of antenna elements enables the end user to point the antenna beam towards a selected direction, maximising the Signal-to-Interference plus Noise ratio (SINR) for the desired data stream, whilst minimising potential co-channel interference.

The application of smart antennas at the system level allows two users to be allocated the same resource as long as they can be spatially separated, being the separation distance dependant on the antenna radiation pattern, as shown by Figure 3. By adjusting the antenna weights to maximise the SINR, two antenna beams can be pointing at the desired co-channel users MSi, and MSj, and will suffer minimal cross talk between beams as long as their angular separation (AoS) θi,j is sufficiently apart; a value dependent on the on the antenna beamwidth. This principle can be applied to both the downlink (DL) and the uplink (UL).

Figure 3- SDMA principle based on angular separation

In order to exploit spatial diversity the resource allocation mechanism evaluates the Mobile Stations (MSs) AoA on every scheduling period resulting in a spatial diversity list. The packet scheduler is a max C/I scheduler that assigns a priority to each packet according to channel strength. Therefore associated with each packet we have 2-samples, i, and σi , where i represents the packet order in the maxC/I list

EESM LUT1

2

1N

SNRSNR

SNR −

⎧⎪⎪⎨⎪⎪⎩

effSNRPER

BS

MSi

MSj

θi,j

236

(where the users are ordered in terms of strength) and σi is the associated AoA.

Let us define the following parameters

• Nt: number of available time slots

• Ns: maximum number of spatially (number of beams supported in the se

• ∆θmin: Minimum required AoS for tlayer to be able to separate two MSs

It is clear that due to the introductidimension, we can increase the number oin one frame from Nt which is the maximslots in the frame with a SISO channel number equal to Ns Nt.

Figure 4 shows the antenna pattern usedThis pattern is obtained theoretically wiantenna elements separated by 0.5 waveleconsists in a 3 dB beamwidth of aboutmaximum side lobe of -13.46 dB. simulations we assume that no co-channefor θi,j ≥ 45 degrees.

Figure 4- Smart Antenna beamform

Figure 5 presents the antenna pattern proa 3-sector cell with a 3 dB beamwidth gain for this antenna is 14 dBi. By reducby half, to 35 degrees, the correspondinghigher resulting in 17 dBi [11]. This isantenna gain used for the SDMA simulati

-25

-20

-15

-10

-5

0

-120 -100 -80 -60 -40 -20 0 20 40

Gai

n in

dB

.

Azimuth in Degrees

3 Sector Antenna Pattern

decreasing channel

separable channels ector)

the PHY (Physical)

ion of the spatial of scheduled packets mum number of time

up to a maximum

d by the transmitter. ith an array of four engths. This pattern t 35 degrees and a Although, in the

el interference exists

ming pattern

oposed by 3GPP for of 70 degrees. The cing the beamwidth g gain will be 3 dB s the corresponding ions.

Figure 5- Antenna pattern f3G

Let us define an array of dimnumber of time slots and Ns The elements of the array wipackets to be scheduled.

Look-Up Table

The interfacing to the link laytables (LUT) simulated usingthe previous section. The transport formats based on thof 1/2, 2/3 and 3/4 using QPwith the respective payloads p

Table 1. Modulation and CodingModulation and

Coding Rate Bl

QPSK, R=1/2 QPSK, R=2/3 QPSK, R=3/4

16QAM, R=1/2 16QAM, R=2/3 16QAM, R=3/4

Figure 6-

Simulation scenario

The simulation parameters aConcerning the traffic modeis characterized by an alwaythis is the commonly adoptecapacity. For the Near Real rate is 2Mbps and is based [12].

Full queue traffic results

Figure 7 shows the average Oand average service sector thrtotal number of bits transmithe service throughput is transmitted bits within the sim

60 80 100 120

n

1.E-03

1.E-02

1.E-01

1.E+00

-4 1 6

FER

SINR

Look-U

for 3 sectors cells proposed by GPP

mension Nt Ns , where Nt is the the number of spatial columns. ill contain packet indices of the

yer is based on a set of look-up g the air interface presented in LUT in Figure 6 covers the he channel coding having rates PSK and 16QAM modulations, presented in Table 1.

g Schemes and respective payloads lock Size (bits)

Max. Bit rate (Mbps)

1536 20.5 2048 27.3 2304 30.7 3072 41.0 4096 54.6 4608 61.4

EESM LUT

are summarized in the Table 2. ls, the full-queue traffic option

ys-full user transmission buffer; ed model for evaluating system Time Video option, the source on the modelling approach in

Over-The-Air (OTA) throughput roughput; where the OTA is the tted over the air interface, and the number of successfully

mulation time [13].

11 16

Up TableQPSK R=1/2QPSK R=2/3QPSK R=3/416QAM R=1/216QAM R=2/316QAM R=3/4

237

Table 2. System-level simulations parameters Parameter Value Environment Urban Mobile velocity 3km/h Frame duration 0.3 ms Channel Model 3 GPP 2 Cell type Sectorized (3 sectors) Cell radius 300 m (size of hexagon)

Traffic models -Full queue (buffer full of data) -NRTV 2Mbps

Users 30 per sector Number of BS Antennas 4

Antenna separation 0.5λ Number slots data 4 Max. of beams per sector

2

Scheduler type Max C/I Link Adaptation BLER ≤ 0.1

Figure 7- Full queue average sector throughput for DRA

with SDMA vs. non-SDMA case

The obtained results have shown that with SDMA, the Over-The-Air (OTA) throughput for the full-queue case can reach near 100 Mbps resulting in a throughout gain of 40 Mbps over non-SDMA DRA case. Lack of diversity gain due to low number of users reduces the simulated SDMA gain. Service Throughput is around 80 Mbps for SDMA and 51 Mbps without SDMA. The discrepancy between service and OTA, although slight, is due to users being serviced at the cell edge which normally experience poor signal quality.

NRTV 2 Mbps results

Figure 8 presents the results of DRA under NRTV traffic for SDMA vs. non-SDMA cases. The use of service traffic models means that the effect of SDMA is less pronounced as in the case of the full queue case, especially for NRTV. Furthermore, we notice about 9 Mbps drop in service throughput compared to the OTA. Using traffic models with a pronounced activity factor and large packet sizes will increase the SDMA cell throughput since users with the highest channel propagation conditions would be continuously served with the highest MCS option (16 QAM, ¾ rate encoder) avoiding underutilization of resources.

Figure 8- NRTV 2Mbps average sector throughput for

SDMA vs. non-SDMA

The attained results lead to conclude that the performance of SDMA is affected by the likelihood of finding a user pair that has an angle of separation greater than the required separation, in this case 45 degrees. User density will affect the amount of multi-user diversity that can be exploited by the scheduler. A low density will reduce the average MCS utilization option (lowest option corresponds to QPSK with ½ rate encoder). A radiation pattern with a lower angular beamwidth will increase the likelihood of finding a user pair within a given coverage area.

4.2. Second Scenario: Using Cooperative Communications

In this section, we tackle the problem of poor wireless channels connecting mobiles (D), located at the cell border for example, to base stations (S) by introducing a relay node between the sources and the mobiles (Figure 9). The relay will connect sources to mobiles and will control the relaying (forwarding) of the packets in a way to adapt to channels conditions and QoS criteria imposed on data types. The main problem of mobiles, located at a far distance from S, is that these mobiles receive a weak signal, due to the poor coverage provided by the base station at the cell border. This weak direct link results in poor performance, longer delays and higher packet error rates.

Our target is to design a basic cooperative MAC protocol that uses information from the physical layer in order to adapt its forwarding process for the mobiles (D). It is a cross-layer approach which optimizes the use of the relay in poor channel conditions. The cooperative communication has two phases. In phase 1, the sources transmit their packets to the relay. In phase 2, the relay that has buffered the packets it had received from the sources can then estimate the channel conditions of the R-D links, as shown in Figure 3. Once estimated, the relay is now willing to use the EESM to estimate the PER at the output of the channel decoder. Using the upper layer information, each mobile data will be assigned by some constraints which reflect the QoS of the transmitted data. In literature, the QoS could be given in terms of latency, throughput, PER or others. In our work, we

0

20

40

60

80

100

120

140

Theoretical OTA Service

Average Cell Throughput (Mbps)

Without SDMA

Using SDMA

0

5

10

15

20

25

30

35

40

OTA Service

Average Cell Throughput (Mbps)

Without SDMA

Using SDMA

238

assume that a target PER, called PERT, is assigned to each D. The problem is now to allocate the resources at the relay node in such a way to verify the target PER. In other words, once the relay receives the estimated PERs of the R-D links, the MAC protocol kicks in and forms a burst of packets with a sharing ratio for each D that depends on their target PER and instantaneous PERs.

Figure 9- cooperative communication at the cell border

We distinguish two main data types, Real-Time (voice and video calls, video and audio streaming, etc) and non-Real-Time (data transfer, web browsing, etc). Each type is associated with a target PER ( and for instance that is sent by the mobiles to the relay. R uses these target PERs in addition to an estimation through EESM technique of the current link’s PERs (Figure 5), to calculate the sharing ratios and allocate resources accordingly. The EESM technique described previously plays therefore a central role as a LSI in this transmission scenario, i.e. cooperative communication at the cell border.

Figure 10- Target and link PERs

Once the parameters and the channels conditions of R-D1 and R-D2 links are obtained, an estimation of the effective SNR value can be obtained and then the PER of the different links can be predicted through EESM. Knowing PER1 and PER2 of different links as well as their respective target PERs, the problem turns out to find the resource allocation algorithm which divides the transmission duration to both destinations according to the required QoS. Using the Lagrange optimisation, we are then able to compute the number of packets allocated to each mobile in one burst of K packets. They are given by:

. .. (6)

where and are respectively the number of packets reserved for the first and second user terminal. The parameter a reflects the PER ratio between both user terminals given by:

(7)

Figure 11 and Figure 12 show that the mobile with a Real-Time data (higher target PER) will get higher throughput (bandwidth) but with higher PER, the other mobile with non-Real-Time data will get lower throughput and lower PER. Of course, we can see that the PER targets are almost verified in all the cases. Moreover, Figure 11 shows that our allocation scheme outperforms the uniform non-adaptive case, by ensuring a lower PER for all values of the parameter (a), this provides less error packets that lead to lower retransmissions and shorter delays.

Figure 11- PER of both mobiles at the destination

Figure 12- Throughput of both mobiles

5. CONCLUSION

In this paper a link-to-system level interface for the simulation of B3G scenarios is presented. The system level platform that aims to assess RRM algorithms is described. The interface to the link-level in OFDM is based on the EESM, which is the most promising technique in such an interface for the proof of concept targets. Two showcases are presented, using SDMA to improve and evaluate the throughput of an OFDM based system, and cooperative communications at cell border.

S SR

D

D

R

M 1 M 2 Target PER 1 Target PER 2

PER 1 PER 2

0 0.5 1 1.5 2 2.5 3 3.5 4

4.4

4.6

4.8

5

5.2

5.4

5.6

5.8x 10

-3

Ratio (a)

PE

R

D1 (Adaptive)D2 (Adaptive)D1 (Uniform)D2 (Uniform)

0 0.5 1 1.5 2 2.5 3 3.5 450

100

150

200

250

300

Ratio (a)

Thr

oug

hpu

t (M

bps

)

D1 (Adaptive)D2 (Adaptive)D1 (Uniform)D2 (Uniform)

239

ACKNOWLEDGEMENTS

This work has been performed in the framework of the ICT projects ICT-217033 WHERE which is partly funded by the European Union, and the NewCom++ project as well.

REFERENCES

[1] ITU-R, “Vision, framework and overall objectives of the future development of IMT-2000 and systems beyond IMT-2000”, draft new recommendation ITU-R M. [imt-vis] [doc. 8/110], Feb. 2003.

[2] IST MATRICE, http://www.ist-MATRICE.org/, IST- 2001 32620

[3] Liang Xu, Xuemin Shen,and Jon W. Mark, Dynamic bandwidth allocation with fair scheduling for WCDMA systems, IEEE Wireless Communications, vol. 9, no. 2, April 2002, pp. 26-32

[4] A. Stamoulis, N. D. Sidiropoulos, G. B. Giannackis, Time-varying fair queueing scheduling for multicode CDMA based on dynamic programming, IEEE transactions on Wireless Communications, vol. 3, Issue 2, March 2004, pp. 512-523

[5] Ratasuk R., Ghosh A., Classon, B., "Quasi-static method for predicting link-level performance", in Proc. of IEEE Vehicular Technology Conference, vol.: 3, pp. 1298-1302, May 2002.

[6] Ericsson, "System level evaluation of OFDM- further considerations", TSG-RAN WG1 #35, Nov. 2003, R1-031303, Lisbon, Portugal.

[7] 3GPP TSG-RAN-1, "TR 25.892: feasibility study for OFDM for UTRAN enhancement", version 1.1.0, March 2004.

[8] P. Liu, Z. Tao, Z. Lin, E. Erkip and S. Panwar "Cooperative wireless communications: a cross-layer approach", in Proc. of IEEE Wireless Communications, Vol. 13, Issue 4, pp.: 84-92, Aug. 2006.

[9] Y. Nasser, J.-F. Hélard, M. Crussière, "System level evaluation of innovative coded MIMO-OFDM systems for broadcasting digital TV", in EURASIP International Journal of Digital Multimedia Broadcasting, vol. 2008, Article ID 359206, 12 pages.

[10] Y. Nasser, M. des Noes, L. Ros, G. Jourdain, "Performance Analysis of OFDM-CDMA systems with Doppler Spread", in Proc. of IEEE Wireless Personal Multimedia Communications (WPMC), vol. 2, pp. 153-157, California, USA.

[11] 3GPP, TR25.996 "Spatial channel model for multiple Input Multiple Output (MIMO) simulations," v6.1.0, 2003-09

[12] 3GPP, TR25.892, "Feasibility Study for Orthogonal Frequency Division Multiplexing (OFDM) for UTRAN enhancement," v6.0.0, 2004-06.

[13] 3GPP, TR25.848 " Physical layer aspects of UTRA High Speed Downlink Packet Access ," v4.0.0, 2001-03

240