Progress In Electromagnetics Research Symposium Proceedings, Moscow, Russia, August 19–23, 2012
1575
A Method of PCI Planning in LTE Based on Genetic Algorithm Hao Sun, Nan Li, Yanlei Chen, Jiangbo Dong, Na Liu, Yunbo Han, and Wei Liu
China Mobile Design Institute, Beijing, China
Abstract— Physical Cell ID (PCI) planning is one of the most important steps in whole Long Term Evolution (LTE) (LTE) network planning and construction. It is difficult to improve the QoS of Networ Network k by some traditiona traditionall methods, methods, such such as antenna antenna adjustment adjustment in LTE system. system. Assigning Assigning PCI reasonably is able to decrease interference and increase operation rate and performance of the network. A PCI planning system is presented in this paper. This system would analyze the interfer interference ence of LTE system and build up an interferenc interferencee matrix. matrix. Then this system would use genetic genetic algorithm algorithm to adjust adjust the distribution distribution of PIC to minimize minimize the interfe interference rence.. Finally Finally a PCI plan scheme would be given by this system. 1. INTRODUCTI INTRODUCTION ON
Physical Cell ID (PCI) is one of the most important cell’s identifier in the wireless network of LTE system. Therefore, PCI planning is one of the most important steps in LTE network planning and construction construction.. To assign and use PCI reasonably reasonably could reduce reduce interfer interference ence and increase increase resource resource utilization and QoS of the LTE system. PCI in LTE only has 168 groups, 504 in total [1], which is not enough for needs of LTE commercial networks. The main idea of genetic algorithms is that adaptable individuals in the group have more chance to hybrid in order to inherit inherit good goo d characteristi characteristics. cs. In this way, way, the group as a whole could adapt to the environmen environment. t. Genetic Genetic algorithm starts with a problem problem which could have have potential potential results. Then, based on the principle of survival of the fittest, the problem would get the best approximate solution solution by the evolution evolution of each generation. generation. In each generation, generation, the most adaptable individuals individuals are picked to produce new population presenting new by combination, crossover and mutation with the genetic genetic operato operators. rs. This procedur proceduree would would lead to evolut evolution ion of the population population to much much more adapta adaptable ble as in the nature. nature. The best indivi individua duall though though decoding decoding in the latest latest populati population on could could seem as the best approximate solution of the problem [2]. LTE network needs not only well Reference Signal Receive Power (RSRP), but also high Signal to Interference plus Noise Ratio (SINR). If PCI could not be planned well, it will cause high interruption of Reference Signal (RS). Then this situation would cause black hole of signal coverage. A method of PCI planning based on genetic algorithm is presented in this paper to assign PCI reasonably for LTE networks. This method could improve the PCI plan result to increase the QoS of LTE network. 2. PRINCIPLES PRINCIPLES
The theoretical theoretical basis of genetic genetic algorithm is schema schema theorem and building building block assumptions assumptions.. The basic follow of genetic algorithm is shown as Figure 1 [3]. Schema theorem has the character of low order, short length and average fitness of the mode of degree higher than the population means fitness in the offspring. A schema could contain many strings, strings, and different different strings strings could link to each other through schema. schema. The string operator in the genetic algorithm is essentially a mode of operation. Statistics researches show that to obtain optimal feasible solution in random search should make sure that the sample of better feasible solutions must grow exponentially. 2.1. Mathematical Mathematical Modeling
There are two ways ways of string and binary encoding in genetic algorithm. algorithm. Binary encoding encoding method is used in this paper. The result of this problem is presented by an n ∗ m two-dimensional matrix F . F . Figure 2 shows that different antenna numbers in LTE system would get different matrix F . F . Every element in this matrix is presented as f ij , which the value is 1 or 0. When PCI j is assigned to cell ij m
i, f ij ij = 1, otherwise f ij ij = 0. In order to satisfy the demand vector, we need to make
f = d , ij ij
j =1
which di is cell i’s PCI demand, n is number of cell, m is the number of usable frequency [4].
i
PIERS Proceedings, Moscow, Russia, August 19–23, 2012
1576
generate the initial population (generate several individual s randomly)
Evaluate fitness functions
Meet optimization criteria
Yes
No
Find the best individual and finish
o G p e e n r e a t t i e c o r
select
cross
mutation
Figure 1: Follow of genetic algorithm.
Figure 2: Mapping of downlink reference signals.
2.2. Initial Population
There are two ways to initial population generally, one is random generation, which is used for the situation of solution to the problem without any prior knowledge, and another one is to transfer some prior knowledge to a group of essential requirements, then choose the population randomly from the results which meet the requirement. In order to get the optimal solution faster, the second way is used to initial population in this paper [5]. 3. SYSTEM REALIZATIONS
Genetic algorithm is packaged in this paper to realize the function of automatic plan for each cell’s PCI in LTE. 3.1. Theory of Automatic PCI Plan
There are several factors needed to consider during PCI code plan, such as PBCH, PDCCH and so on. In this system, user could choose one or multiple channels as the considerations during PCI plan. As known, the different channel has different parameter. For example, PBCH uses initial parameter Cell C Init = N ID , (1) Cell which N ID is PCI of this cell. PDCCH uses initial parameter Cell C Init = ns /2 ∗ 29 + N ID ,
(2)
Cell which ns is number of timeslot in frame, N ID is PCI of this cell. PCFICH uses initial parameter
C Init
= (n /2 + 1) ∗ 2 ∗ N
Cell ID
s
+ 1 ∗ 2 + N
Cell ID ,
9
(3)
Cell which ns is number of timeslot in frame, N ID is PCI of this cell.
Correlation of scrambling code in LTE is calculated by Rab = of LTE system, an improved method is used as follows: Rab
1 = N
N
a i=1
M
· bi ,
M =
N,
1
N
N
a · b . Considering the delay i
i
i=1
(i + m) mod N = 0 (i + m) mod N, (i + m) mod N =0
(4)
Progress In Electromagnetics Research Symposium Proceedings, Moscow, Russia, August 19–23, 2012
1577
start
1. select cells
2.generate PCI plan requirement list
input
P s y C s I t e p l m a n
3. select usable PCI PCI plan requirement
Interference matrix
4. generate cell interference matrix Database 5. initial populate
PCI plan (genetic algorithm )
output 6. evaluation
solution
10. mutation 9.cross 8. selection
Figure 3: Construction of PCI plan system.
Is get the best approximate solution No 7. Save the best individual, or replace the worst individual
Yes
11. adjustment 12. output
finish
Figure 4: Flow of PCI plan algorithm.
3.2. PCI Automatic Plan
PCI plan module is used for assisting LTE network plan and optimizes and gets best solution of PCI plan. System would obtain cell information, such as frequency, and measurement report, etc. before PCI planning. Then system would assign PCI for cells in order to reduce the interference of the whole LTE network as far as possible. At last, the system would output the solution for users. PCI plan module is shown in Figure 3, which, the flow of PCI plan algorithm is shown in Figure 4. 4. SIMULATION RESULT OF PCI PLAN
All the algorithm and method introduced above are coded with C# and used in LTE plan tool ANPOP. This PCI plan system also needs Geographic Information System (GIS) as an addition condition. In this section, we use RS and PBCH’s SINR to evaluate the effect of PCI plan method based on genetic algorithm by compare with the simulation result of manual and automatic PCI plan. At the same time, Monte Carlo (MC) simulation is also launched in this paper. Assuming there are 5 UEs every cell, service type is Full Buffer FTP. The result of MC shows as follows: Manual PCI plan, average cell throughput is 16.3 Mbps, average cell edge throughput is 0.43 Mbps; Automatic PCI plan using genetic algorithm, average cell throughput is 20.1 Mbps, average cell edge throughput is 0.65 Mbps. To compare the two results, we could know that, using the second method could increase cell throughput 23.3%, and cell edge throughput 51.2%. According to the results above, the algorithm mentioned in this paper could reduce same frequency disturbance and improve SINR coverage of the network, and also could increase the throughput of the LTE network.
PIERS Proceedings, Moscow, Russia, August 19–23, 2012
1578
RS simulation result by PCI plan manually
RS simulation result by PCI plan of genetic algorithm
RS SINR
Area (km 2)
Persentage (%)
− 15.0 =< X < − 10.0
0.025
0.1
− 15.0 =< X < − 10.0
0.000
0.0
− 10.0 =< X < − 3.0
4.988
13.6
− 10.0 =< X < − 3.0
0.025
0.1
− 3.0 =< X < 5.0
4.187
11.4
− 3.0 =< X < 5.0
4.988
13.6
5.0 =< X < 10.0
8.137
22.2
5.0 =< X < 10.0
4.187
11.4
10.0 =< X < 20.0
15.330
41.8
10.0 =< X < 20.0
8.137
22.2
20.0 =< X < 40.0
4.016
10.9
20.0 =< X < 40.0
15.330
41.8
40.0 =< X
0.000
0.0
40.0 =< X
4.016
10.9
RS SINR
Area (km 2)
Persentage (%)
Figure 5: Simulation results of RS’s SINR. 5. CONCLUSIONS
From the simulation result above, we could get the conclusion that assigning PCI reasonably is able to effectively reduce the interference in LTE system, especially in the case of constructing network using the same frequency, and to improve coverage and QoS of the network. As a result, PCI plan algorithm presented in this paper is suitable for multi-antenna LTE system. ACKNOWLEDGMENT
Especially thanks to my lovely wife; I cannot finish this paper without her support. REFERENCES
1. 3GPP, “Evolved universal terrestrial radio access (E-UTRA); Physical channels and modulation (release 8),” 3GPP TS 36.211, 72–74, Dec. 2009. 2. Shen, J. and S. Suo, 3GPP Long Term Evolution (LTE) Technical Principle and System Design , Posts & Telecom Press, Beijing, 2010. 3. Ngo, C. Y. and V. O. K. Li, “Fixed channel assignment in cellular radio networks using a modified genetic algorithm,” IEEE Transactions on Vehicular Technology , Vol. 47, No. 1, 163– 172, 1998. 4. 3GPP, “Evolved universal terrestrial radio access (E-UTRA); Radio resource control (release 8),” 3GPP TS 36.331, 75–76, Dec. 2010. 5. Wang, X. P. and L. M. Cao, Genetic Algorithm — Theory, Application and Software Implementation , Xi’an Jiaotong University Press, Xi’an, 2003.