a cell planning scheme for wcdma systems using genetic

23
A Cell Planning Scheme for WCDMA Systems Using Genetic Algorithm and Measured Background Noise Floor Hsin-Piao Lin 1 , Rong-Terng Juang 1 , Ding-Bing Lin 1 , Cheng-Yi Ke 2 and Yi Wang 2 1 Institute of Computer, Communication and Control, National Taipei University of Technology. No. 1, Sec. 3, Chung-Hsiao E. Road, Taipei, Taiwan. [email protected], [email protected], [email protected] 2 Taiwan Cellular Co., 4Fl.-2, No. 10, Lane 609, Sec. 5, Chung-Shin Rd., San-Chung, Taipei, Taiwan. Abstract –WCDMA is an interference-limited system with coverage and data throughput sensitive to background noise. This paper presents the background noise measurements in urban Taipei city for the licenses bands of 3G systems issued in Taiwan. The measurements involve with FDD mode uplink and downlink frequency bands measured on building tops and at street level, respectively. The severeness of spectrum pollution of these bands is evaluated by extracting three statistic parameters from the measurements, and the impact of the background noise on coverage and throughput is analyzed for the WCDMA systems. Besides, basing on measurements, a better solution using a genetic algorithm with the help of propagation model and digitized building information is proposed for the cell planning of WCDMA systems, by which the required coverage can be met with the optimum solution for BS number, locations, antennas heights, and transmitting power, so as to obtain a system suffering from less impact of the background noise and achieve higher data throughput with minimum cost. Keywords – WCDMA, background noise, cell planning, genetic algorithm.

Upload: toh-pyie-shein

Post on 13-May-2017

219 views

Category:

Documents


2 download

TRANSCRIPT

A Cell Planning Scheme for WCDMA Systems Using Genetic

Algorithm and Measured Background Noise Floor

Hsin-Piao Lin1, Rong-Terng Juang1, Ding-Bing Lin1, Cheng-Yi Ke2 and Yi Wang2

1Institute of Computer, Communication and Control, National Taipei University of Technology.

No. 1, Sec. 3, Chung-Hsiao E. Road, Taipei, Taiwan.

[email protected], [email protected], [email protected]

2 Taiwan Cellular Co., 4Fl. -2, No. 10, Lane 609, Sec. 5, Chung-Shin Rd., San-Chung, Taipei,

Taiwan.

Abstract–WCDMA is an interference-limited system with coverage and data

throughput sensitive to background noise. This paper presents the background

noise measurements in urban Taipei city for the licenses bands of 3G systems

issued in Taiwan. The measurements involve with FDD mode uplink and downlink

frequency bands measured on building tops and at street level, respectively. The

severeness of spectrum pollution of these bands is evaluated by extracting three

statistic parameters from the measurements, and the impact of the background

noise on coverage and throughput is analyzed for the WCDMA systems. Besides,

basing on measurements, a better solution using a genetic algorithm with the help

of propagation model and digitized building information is proposed for the cell

planning of WCDMA systems, by which the required coverage can be met with the

optimum solution for BS number, locations, antennas heights, and transmitting

power, so as to obtain a system suffering from less impact of the background noise

and achieve higher data throughput with minimum cost.

Keywords – WCDMA, background noise, cell planning, genetic algorithm.

1. INTRODUCTION

Third generation (3G) mobile communication systems brought up many attentions in

these few years. Many governments have released licensed frequency bands for 3G

services. Among the proposed radio transmission technologies, WCDMA (Wideband

code division multiple access) technology has emerged as the most widely adopted 3G

air interface. CDMA is characterized as an interference-limited system, i.e. an increase

of the interference level determines a decrease of the system performance. The

performance degradation due to interference from adjacent narrow band systems has

been evaluated for a WCDMA network [1][2], where the capacity per cell is sensitive to

the coverage and interference.

In Taiwan, the 3G bands are divided into five licenses. Due to the tradeoff between

the cost and performance, system operators need to evaluate the severeness of the

spectrum pollution of those bands before deploying the networks. A direct solution is

conducting the measurements for the background noise floor and utilizing statistic

parameters to process the large number of measurements. Meanwhile, these

measurement data could be used for cell planning in the initial stage of system

development.

Cell planning, the key the system performance and economical efficiency, is a

complex and important issue in cellular communication systems. For the degrading of

system performance due to interference in WCDMA networks, cell planning could not

base only on signal prediction but must also consider the power limits and the signal

quality constraints. A mathematical programming model, proposed in [3], considers the

Signal-to-Interference Ratio (SIR) to support the decisions on where to locate the new

base stations (BS) and which configuration to be selected for each of them.

Numerous attempts have been proposed to optimize network performances in terms

of capacity, coverage, quality of service, etc. A tuba search approach for cell planning

with capacity expansion was proposed in [4], which considers two types of BSs: some

existing ones in service and some additional BSs needed to be determine for the

increased traffic demand. The influence of site location and antenna tilts onto the

operation of UMTS systems were presented in [5]. A genetic algorithm (GA) based

automatic cell planner was proposed in [6], which adjusts antenna parameters and BS

transmitted power to improves the network performance. The optimizing of BS

locations as well as their configurations was addressed in [7], which utilizes a

mathematical programming model considering the power control mechanism and the

SIR as quality measures and employs a tabu algorithm to find the approximate solutions

of the problem.

However, little attention has given to the impact on background noise on system

performance. This paper proposes a methodology for the initial development of

WCDMA systems to mitigate the influence of the background noise and increase the

throughput. A GA with the help of Walfisch-Ikegami propagation model, verified as an

accurate model for predicting propagation path loss in urban area with smaller cells [8],

and digitized building information is used to achieve the optimization of cell planning.

GA, developed by Holland [9], is a nature-inspired algorithmic technique basing on the

principles of natural evolution and widely used to solve optimi zation problems [10][11].

In the proposed method, the required coverage can be met with the optimum solution

for BS numbers, locations, and antennas heights.

This paper is organized as follows. Section 2 presents the measurements of

background noise level in urban Taipei city. Statistical parameters are used to evaluate

the spectrum qualities of the 3G license bands. The performa nce degradation due to the

background noise for the WCDMA system is also studied here. Section 3 addresses the

cell planning methodology based on the background noise measurements. Section IV

takes an example for deploying a WCDMA network in a real environment. In Section V,

conclusions are drawn.

2. BACKGROUND NOISE FLOOR MEASUREMENTS IN URBAN TAIPEI

CITY

Table 1 summarizes the frequency bands of 3G services issued in Taiwan. Except

license E, each one has a FDD (frequency division duplex) mode uplink frequency band,

a FDD mode downlink band, and a TDD (time division duplex) mode frequency band.

During the summer of 2001, we carried out lots of background noise measurements

involving with FDD mode uplink and downlink bands in urban Taipei city. The

measurements for uplink and downlink bands were conducted on building tops and at

street levels, respectively. A drive test solution for WCDMA, Agilent E7476A, was use

to complete the measurements for the downlink frequency bands, and a spectrum

analyzer, ADVENTEST U3641, was used to measure the noise for the uplink bands at

four directions, east, west, south, and north at each selected location. Licenses A, B, C

and D were measured for both uplink and downlink bands, and the frequencies, power

levels and GPS coordinates were recorded in a notebook during the measurements.

Three statistic parameters are used to process the large number of measurements for

the evaluations of the spectrum clearness. The first parameter is the cumulative

distribution function (CDF) of noise power levels, which reveals the percentage of noise

levels under a certain power thresholds. The second one is the statistics of frequency

domain power level crossing rate (FD-LCR), the rate that the noise power envelope

crosses certain specified power thresholds in a positive-going direction. For simplicity,

this quantity is taken as the average of the rates obtained from the measurement at the

four directions. However, CDF and FD-LCR are not sufficient for spectrum quality

evaluation because the noise bandwidth is a more important dominator for

communication quality in CDMA systems. Hence, the third parameter is the average

noise bandwidth, in which the noise power levels are above certain specified power

thresholds. Considering the FD-LCR and average noise bandwidth together, by which

gives the noise bandwidth and crossing times per unit bandwidth at different power

thresholds, it is easily to distinguish the relative spectrum qualities between the bands.

Figure 1, 2, and 3 are the statistical information of the license bands measured at the

same location, where (a)s are that for the FDD mode uplink bands and (b)s for the

downlink bands. Figure 1 shows the CDFs of the background noise power levels, and

the severeness of spectrum pollution in ascending order is license D, C, B then A in

terms of the CDFs. Figure 2 exhibits the average FD-LCRs of the background noise

envelopes measured at the four directions, and the severeness in ascending order is also

license D, C, B then A in terms of the FD-LCRs. Figure 3 displays the average

bandwidth of the background noise. Though a slight ambiguity appears in the

distinguishing of pollution severeness, it is confident to conclude the spectrum quality

in descending order as license D, C, B then A.

Having estimated the spectrum qualities, the discussion goes on to the impact of the

background noise levels on cell coverage and throughput for WCDMA systems. The

analysis begins by contemplating the definition of the required signal to noise ratio

(SNR). The uplink case is explained in the first place and the SNR of user j is given by

[12]

)(

)(

cjNF

Rxj

jjj IP

P

RvW

SNR+

⋅= (1)

where W is the chip rate, 3.84Mcps, jv is the activity factor at physical layer, 0.67 for

speech and 1.0 for data, jR is the bit rate, )(RxjP is the received power from user j,

NFP is the background noise floor including thermal noise and noise from any

man-made transmitters within other communication systems, and )(cjI , the co-channel

interference associated with user j, is defined as the interference power coming from

other links operating at the same frequency band within the same system. The total

received power totalI can be expressed as the summation of the background noise floor,

co-channel interference, and the received power from user j, i.e.

)()( Rxj

cjNFtotal PIPI ++= . Defining totalj

Rxj ILP ⋅=)( , the load factor jL of one link has

the form

jjjb

j

vRNEW

L

⋅⋅+

=

)/(1

1

0

(2)

where jb NE )/( 0 is the SNR of link j. For N users with the same traffic type in the cell,

the system loading η is

∑=

=N

jjL

1

η (3)

The received power excluding the background noise floor can be given by

totalNFtotal IPI ⋅=− η (4)

It is more precise to consider each single service (12.2kpbs, 144kpbs, etc.) in system

dimensioning at each time. However, for succinctness, a compound traffic pattern is set

as a mixed of 80% speech users (12.2kpbs user data rate), 15% of 144kpbs and 5% of

384kpbs data users in the performance analyses below. The minimum SNR

requirements for the traffic pattern are assumed as 5dB, 1.5dB and 1.0dB, respectively.

According to (2), the averaged load factor, jL , of one link with respect to the specified

traffic pattern is 0.0185 if the SNR of every link reaches the minimal signal, which

leads to the acceptable BER (bit error rate) performance. In the calculation of uplink

coverage, the total received power, totalI , can be obtained using (4) if the system

loading and the background noise floor are given. Afterward the minimum received

power from user j is determined by totaljRx

j ILP ⋅=)( . Given the mobile transmitting

power as 0.5W, the maximum allowed path loss can be evaluated because )(RxjP is the

product of the mobile transmitting power multiplying the path gain, the path loss

expressed in decibel. Fig. 4 shows the relationship between the background noise floor

and the maximum allowed uplink path loss under different system loadings, 50%, 60%,

and 70%. The figure indicates that a lower noise level determines a larger cell coverage

range.

The analysis of the downlink throughput is based on a similar principle as the uplink

case. The BS transmitting power is 1W, the compound traffic pattern is adopted, and 4

different maximum allowed path losses, 130dB, 135dB, 140dB and 145dB, are set. Here,

the SNR of each link is also asked to meet the minimum requirement. Accordingly, the

total received power totalI is solved by totaljRx

j ILP ⋅=)( . Once the background noise

floor is given, the system loading η can be obtained according to (4). Consequently,

the system throughput is yield by the product of the average data rate multiplying the

system loading. Figure 5 shows the relationship between the background noise floor and

maximum downlink throughput. This figure exhibits that the throughput is higher at a

lower noise level.

Table 2 summarizes an example of the system performance evaluation from the

measurements for the uplink and downlink bands of license A, B, C and D. The

transmitting powers of the BS and mobile are set as 1W and 0.5W, respectively, and the

traffic pattern is the same as previous studies. The table shows that license D is the band

with the lowest mean noise power and maximum cell range for uplink, and license A

suffers from heavier impact of background noise. For downlink, license D is the one

with lower mean noise power, and license C suffers from heavier impact of background

noise.

3. CELL PLANNING USING GENETIC ALGORITHM

The proposed cell planning scheme focuses on the mitigation of performance impact

from background noise and efficiently operating of the WCDMA networks. Basing on

the background noise measurement, a simple GA with the help of Walfisch-Ikegami

model and digitized building information is proposed to achieve the optimization of cell

planning for the WCDMA system. The detailed description is given below.

A. Simple Genetic Algorithm

GA is a nature-inspired algorithmic technique for optimization problems based on the

principles of natural evolution. The individuals with better gene, leading to be fitter for

the environment, survive in the evolution process, but otherwise eliminated. After the

elimination, the survivals mate with each other and bear their offspring. The offspring

inherit their parents’ genes, which are the same as their parents or even better.

Consequently, the best gene could be obtained by iterating the evolution process.

The GA used here is a binary version and much likely as in [13]. The following gives

a briefly description of the main components used in binary GA;

1) Chromosome Encoding: The GA begins by defining a chromosome, which is

encoded as a binary string according to the characteristics of each individual.

2) Fitness Function: It is used to evaluate the fitness of every individual in the

environment. The user must decide which parameters of the problem are most

important because too many parameters bog down the GA.

3) Selection: It occurs at each generations or iteration of the algorithm. The individuals

with higher fitness will survive for mating, but otherwise will be discarded to make

room for the new offspring.

4) Crossover: It is the creation of offspring by recombining the chromosomes of

selected parents. Basic crossover methods include One-point crossover, multi-point

crossover, and uniform crossover [11].

5) Mutation: It introduces traits not in the original population. For binary operation, the

mutation is implemented by changing 1s to 0s or 0s to 1s on certain randomly

selected points in chromosomes.

Basing on the standard operations as selection, crossover and mutation, GA can solve

optimization problem easily.

B. Cell Planning Using Simple Genetic Algorithm

Figure 6 is the flowchart of the proposed cell planning. The possible locations for

setting BSs are selected according to the digitized building information, and then the

performance of each BS is evaluated basing on the principles in section 2. The

maximum uplink allowed path loss is determined under the conditions 1) 0.5W mobile

transmitting power, 2) 75% system loading, and 3) the presence of uplink background

noise in the cell. The background noise floor used here is the average of measurements

at street level for downlink case and on building tops for uplink case within a radius of

300m. The cell range is calculated by Walfisch-Ikegami model [14][15], which is a

hybrid model combining with diffraction down to street level and some empirical

correction factors. The throughput is evaluated under the conditions 1) specified

transmitting power, 2) the cell’s maximum uplink allowed path loss, and 3) the presence

of downlink background noise in the cell. Accordingly, the possible BSs’ locations

accompanied by their cell performance are obtained, and the next step goes to utilize the

simple GA to decide what combination of these possible BSs is the best solution to cell

planning. The decision is based on the fitness of each chromosome using the fitness

function, which considers the efficiencies of coverage and transmission and designed

for chromosome i as

iTi

Ti

ref

avgi

ii BCCCC

TT

CF1)0,max(

2)(

−−

⋅+= (5)

where iC is the percentage of the covered area, )(avgT is the average data throughput,

avgP is the average transmitting power of used BSs, refT is the throughputs with

respect to the measured minimum noise power level, TC is the desired coverage, and

B is the number of used BSs. The survivals are selected for mating in Selection and

generate offspring in Crossover. The last operation is mutation for introducing the traits

not in the population. The iteration stops to output the solution if the algorithm

continues without improvement, i.e., the algorithm continues the same best

chromosome for a certain iterations.

4. AN EXAMPLE OF DEPLOYING WCDMA CELLS

The selected area for simulation and validation is in the vicinity of NTUT (National

Taipei University of Technology) campus as shown in Fig. 7, which is an area of 2.5kms

by 1.6kms square digitized building map. Those blocks with brighter color represent

higher buildings and darker ones represent lower buildings. The average building height

and the standard deviation are 18.2m and 13.6m, respectively. Four BS transmitting

powers, 1W, 1.5W, 2W and 2.5W, are available, and the mobile transmission power of

0.5W and the same compound traffic pattern are set. The first step of the planning is the

selection of the possible locations for setting BSs according to the digitized building

information. One of the selection criterions is that the BS antenna height must higher

than average building height and lower than 43m. The second step is the evaluation of

each BS’s cell performance impacted by the measurement background noise. Finally,

the characteristics of these possible BSs are encoded as chromosomes and started

evolution by the GA. Here, uniform crossover method is used, and the desired coverage,

the survival rate, and the mutation probability are set as 90%, 50%, and 0.15,

respectively. Each generation composes of 200 chromosomes, and the iteration is

stopped to output the solution when the algorithm continues with the same best

chromosome for 50 iterations. Figure 7 shows the simulation results, where there BSs

with omni-directional antennas (⊕), as marked as BS1, 2 and 3, are needed for serving

this area. Their transmitting powers are 2.5 W, 2.0 W and 1.0 W, respectively, and

antenna heights are the same as 43m. The dotted points are the area covered by the

designed WCDMA system and the coverage rate is 92.6% with total throughput as 7.52

Mbps.

5. CONCLUSIONS

The coverage and throughput of 3G systems are sensitive to noise power level. This

paper has presented the results of noise power measurements in urban Taipei city for the

3G license bands issued in Taiwan. The severeness of the spectrum pollution of these

bands is evaluated by extracting some statistic parameters from the measurements. The

impact of background noise on coverage and throughput for WCDMA systems has been

analyzed, and the spectrum qualities for uplink and downlink bands of license A, B, C

and D have been evaluated. Also, basing on the noise measurements, a cell planning

scheme is developed using a simple GA with the help of Walfisch-Ikegami model and

digitized building information. A selected example shows that the proposed method can

reduce the impact of background noise and use minimum BSs to achieve the maximum

coverage and throughput. Thus, an easy and efficient cell planning for WCDMA

systems in the initial stage of system development can be delivered, and the deployed

system would suffer from less impact of background noise power and achieve

maximum performance with minimum cost. For an extended discussion, the

considerations should include traffic distribution, interference prevention, cell

sectorization, etc, when the amount of the subscribers increases. The growth of traffic

yields an increase of interference and a complicated cell planning. The cost function for

system optimization should be elaborately designed due to the multifarious

interconnected criteria. The discussions of the issues as a whole are taken as the future

work.

ACKNOWLEDGMENTS

This research is sponsored by Taiwan Cellular Co., Taiwan, under Contract

PSCF-91-006.

REFERENCES

[1] B. Smida, V. Sampath and P. Marinier, “Capacity degradation due to coexistence

between second generation and 3G/WCDMA systems,” in Proc. IEEE Veh. Technol.

Conf., vol. 1, May 2002, pp. 95-99.

[2] K. Heiska, H. Posti, P. Muszynski, P. Aikio, J. Numminen and M. Hamalainen,

“Capacity Reduction of WCDMA Downlink in the Presence of Interference From

Adjacent Narrow-Band System,” IEEE Trans. on Veh. Technol., vol. 51, Issue 1, pp.

37-51, Jan. 2002.

[3] E. Amaldi, A. Capone and F. Malucelli, “Optimizing Base Station Siting in UMTS

Networks,” in Proc. IEEE Veh. Technol. Conf., vol. 4, May 2001, pp. 2828-2832.

[4] C. Y. Lee and H. G. Kang, “Cell Planning with Capacity Expansion in Mobile

Communications: A Tabu Search Approach,” IEEE Trans. Veh. Technol., vol. 49, no.

5, pp. 1678-1691, Sep. 2000.

[5] M. J. Nawrocki, and T. W. Wieckowski, “Optimal site and antenna location for

UMTS output results of 3G network simulation software,” in Proc. Int’l Conf.

Microwaves, Radar and Wireless Commun., vol. 3, May 2002, pp. 890-893.

[6] Z. Altman, J. M. Picard, S. Ben Jamaa, B. Fourestie, A. Caminada, T. Dony, T, J.F.

Morlier and S. Mourniac, “New challenges in automatic cell planning of UMTS

networks,” in Proc. IEEE Veh. Technol. Conf., vol. 2, Sept. 2002, pp. 951-954.

[7] E. Amaldi, A. Capone, E. Malucelli, and F. Signori, “UMTS radio planning:

optimizing base station configuration,” in Proc. IEEE Veh. Technol. Conf., vol. 2,

Sept. 2002, pp. 768-772.

[8] Hsin-Piao Lin, Ding-Bing Lin and Rong-Terng Juang, “Performance Enhancement

for Microcell Planning Using Simple Genetic Algorithm”, in Proc. IEEE Antennas

and Propagat. Society Int’l Symp., Vol. 4, Jun. 2002, pp. 664-667.

[9] J. H. Holland, “Adaptation in Natural and Artificial Systems,” Ann Arbor: Univ. of

Michigan Press, 1975.

[10] D. E. Goldberg, “Genetic Algorithms in Search, Optimization, and Machine

Learning,” Reading, Mass.: Addison-Wesley, 1989.

[11] R. L. Haupt, and S. E. Haupt, “Practical Genetic Algorigth,” John Wiley & Sons,

1998.

[12] Harri Holma and Antti Toskala, “WCDMA for UMTS— Radio Access for Third

Generation Mobile Communications,” John Wiley & Sons. 2001.

[13] P. Calegari, F. Guidec, P. Kuonen, and D. Wagner, “Genetic Approach to Radio

Network Optimization for Mobile Systems,” in Proc. IEEE Veh. Tech. Conf., vol. 2,

May 1997, pp. 755-759.

[14] S.R. Saunders, “Antennas and propagation for wireless communication systems,”

John Wiley & Sons, 1999.

[15] D. Har, A. M. Watson, and A.G. Chadeny, “Comment on diffraction loss of

rooftop-to-street in cost 231 Walfisch-Ikegami model,” IEEE Tran. on Veh.

Technol., vol. 48, pp. 1451-1452, 1999.

Table 1

The Spectrum Distribution of The 3G Frequency Bands Issued in Taiwan

FDD mode

uplink downlink TDD mode

License A 1920~1935 MHz 2110~2125 MHz 1915~1920 MHz

License B 1935~1945 MHz 2125~2135 MHz 2010~2015 MHz

License C 1945~1960 MHz 2135~2150 MHz 2015~2020 MHz

License D 1960~1975 MHz 2150~2165 MHz 2020~2025 MHz

License E 825~845 MHz 870~890 MHz

Table 2

An Example of Performance Evaluation of the 3G License Bands Issued in Taiwan

License A License B License C License D FDD

UL FDD DL

FDD UL

FDD DL

FDD UL

FDD DL

FDD UL

FDD DL

50% system loading

149.4 151.8 150.1 151.5 150.7 150.9 150.9 152.5

60% system loading

148.4 150.8 149.1 150.5 149.7 149.9 149.9 151.5

Max. Allowed

Path Loss (dB) 70% system

loading 147.1 149.5 147.8 149.2 148.4 148.6 148.6 150.2

Lmax=130 2.72 2.73 2.72 2.73 2.72 2.72 2.72 2.73

Lmax=134 2.71 2.72 2.71 2.71 2.71 2.71 2.71 2.72

Lmax=140 2.65 2.69 2.66 2.68 2.67 2.67 2.67 2.69

Max. Throughputs

(Mbps)

Lmax=145 2.48 2.59 2.52 2.58 2.55 2.55 2.55 2.61

UL:Uplink DL:Downlink

Fig. 1. CDFs of the background noise power levels. a) Uplink frequency bands and b)

downlink frequency bands.

Fig. 2. FD-LCRs of the background noise envelopes at different power thresholds. a)

Uplink frequency bands and b) downlink frequency bands.

Fig. 3. Average bandwidth of the background noise at different power thresholds. a)

Uplink frequency bands and b) downlink frequency bands.

Fig. 4. Maximum allowed path loss at different background noise power floors for

different system loadings in the uplink WCDMA system.

Fig. 5. Maximum data throughput at different background noise power levels for

different path loss (Lmax) in the downlink WCDMA system.

Fig. 6. Flowchart of the proposed cell planning algorithm.

Fig. 7. A cell planning example. Three BSs cover the dotted points.