You are on page 1of 4

A Dynamic Neighbor Cell List Generating Algorithm in Cellular System

Huaying Zhou Southwest Electronics & Telecom Technology Research Institute,


AbstractA suitable and dynamic neighbor cell list will increase the soft handoff success rate thus decrease the drop probability. Thus it is significant to design an effective neighbor cell list generating algorithm. The conventional algorithms generate the list by simply putting all the neighbor cells together. That will increase the pilot search time, so the handoff success rate reduces. This paper presents a dynamic neighbor cell list generating algorithm based on vote model. Besides, it is optimized by using the handoff statistical information to make the vote-coefficient adaptive by itself. Keywords-component; neighbor cell list; dynamic; vote; votecoefficient

Chengdu 610041, China Email: smilewine@126.com


However, it may not solve the neighbor cell list build problem just accept and reject the cells by sorting the pilot strength of the Active Sets. In this paper, an algorithm of dynamic neighbor cell list planning is provided, in which the vote model is used to solve the neighbor cell acceptance and rejection problem. In section 2, the algorithm model is described first, then a neighbor cell build algorithm based on this model and an improved one using statistical information are proposed. Finally, section 4 summarizes chief contributions of this paper and gives the problems to be studied. II. DYNAMIC NEIGHBOR CELL LIST BULID ALGORITHM

I.

INTRODUCTION

The 3rd mobile communication systems, such as cdma2000, WCDMA, adopt the Mobile Assistant Handoff (MAHO) method. In this method the mobile measures the pilot strength of the neighbor cells and compares them with the pilot strength of the cell in use. If the pilot strength of the neighbor cells is higher than that in use, the mobile reports this result to the base station for the handoff decision-making. Pay attention that in this process the mobile only measures those pilots in the neighbor cell list which the base station notified. Thus if the neighbor cell list built by the base station is incomplete, the mobile can not find those stronger pilots. Unfortunately, these stronger pilots will be strong disturbers toward CDMA systems. It is obvious that a suitable and dynamic neighbor cell list will increase the soft handoff success rate thus decrease the drop probability. That is the reason why it is so significant to research a better neighbor list generating algorithm. Two neighbor list generating algorithm are commonly used in nowadays CDMA systems. A simple algorithm generates the neighbor cell list by inserting all of the cells bordering upon the cell which pilot are in the Active Set. This algorithm only adapts to the small system because upon a large-scale system it will generate a very large neighbor cell list which will highly increase the mobile searching time. In an improved algorithm, the stronger pilot takes priority of inserting its neighbor cells into the list after sorting all the pilots in the Active Sets by their strength. While the number of neighbor cells in the list is reasonable, the neighbor cells of those weaker pilots can not be added in the list and will be discarded. This improved algorithm is more reasonable than the simple one by considering the difference of the pilot strength which reflecting the probability that the mobile moves into neighbor cells.

A. Outline A reasonable neighbor cell list should include the cells which the mobile will possibly move into, thus it is actually a problem that forecast the direction which the mobile move to. The pilot strength is the crucial factor in a handoff decisionmaking process for it indicating the signal quality which the mobile can acquire and the probability of the mobile communicate with each cells at a certain extent. Consequently, it is reasonable to build the neighbor cell list using the strength information of pilots in Active Set. The algorithm presented in the paper not only considers the pilot strength information, but also introduces the weighted average and statistic average. That make the neighbor cell list built even more accord with actual instance. B. Algorithm prototype The primary idea of the model is that each member in set A votes its own neighbor cells for being added into the final neighbor cell list according to its pilot strength firstly, then all of the elected cells will be sorted by their votes. The details are as follows: Members in set A (1, 2N) can take part in voting and the voting strength are decided by their characteristics respectively, Each i can vote 1i2i pi at the same time. Every selected qn obtains the same votes. If qn =qm, that means qn is elected by m as well as n, so the final votes that qn (i.e. qm) obtained is the sum of the votes which n and the m given.

978-1-4244-3693-4/09/$25.00 2009 IEEE

Sort all of the selected11, 21p1, 12, 22p2 1N, 2NpN by their votes obtained. Obviously, the1i, 2ipi selected by i can obtain more votes if the i has greater voting strength, so these objects have priority in being sorted. Besides, the qn has more votes and priority in being sorted if it selected not only by n but also by m. Then take the first M cells in the sorted objects according to actual requirement. A vote process is accomplished C. Neighbor Cell List build algorithm This paper presents a neighbor cell list build algorithm .The details are as below: The base station receives a request of the mobile to perform a handoff and gets a new Active Set. Here the new Active Set can be regarded as the set A in the model and each pilot in the Active Set namely i has vote qualification. The voting strength of a pilot, namely the votes a pilot can give, is a function of its strength.

pilots than only by one pilot. That is just the conclusion the algorithm presented in this paper reached. D. Algorithm Optimization The algorithm can be optimized by taking the statistical information of the actual system. Described previously, the votes that the pilot in Active Set can give the cells actually bordering upon is the same, namely W i=f (S i). Whereas in the actual system the probability of the mobile moving into differ neighbor cells is also different owing to the influence of path and barrier. S.Magnusson and H.Olofsson proposed to build the neighbor cell list considering the statistical information [1] [2]. If the times proportion that mobiles have completed handoff from cell A to cell B or from cell A to all neighbor cells exceeds a certain gate, cell B should be set as a neighbor cell of cell A. Take this idea into account, the algorithm presented in this paper is more advanced. The base station calculates the times Nij that mobiles have performed handoff from a cell (pilot is i) to another cell (pilot is j). Then the vote strength Wij that pilot i can vote for pilot j, namely the votes i can give j, is defined.

Wi = f (S i )

Here the f (.) is an increasing function of the pilot strength. In this paper, it is defined as follows:

f (S i ) = Si + BASE

N ij Wij = g N total

f (S i )

Here BASE is an important parameter of this function and how it is set will highly decides the performance of the algorithm. That will be analyzed in Section 4 by the emulate results. Each pilot (i) in the Active Set all votes for its neighbor cells, such as 1i2i pi. The same pilot i gives its every neighbor cells equal votes, i.e. W i=f (S i). Calculate the number of the votes which all the selected cells separately. For certain cell K may be a neighbor cell of pilot i as well as pilot j, the votes should be the sum of the votes all the pilots have given.

Here Ntotal is the total times that mobiles have performed handoff from the cell (pilot is i) to its neighbor cells. The improved algorithm is more applicable for the actual system. For example, mobiles rarely perform hand off from cell A to cell B due to some nature cause although cell A borders cell B in fact. Thus in the algorithm not optimized, cell B will still be set as a neighbor cell of cell A. However, in the improved algorithm, cell B will not be regarded as a neighbor of cell A for g (NAB/Ntotal) is small owing to few handoff times for cell A to cell B (NAB ). Besides the virtues mentioned before, the optimized algorithm will automatically adjust the vote strength according to the statistical information. If mobiles in cell A performed handoff more frequently to cell B evidently in a certain period such as on and off duty, the vote strength that cell A can give to cell B will be increased due to the increasing of g (NAB/Ntotal). That will increase the priority of cell B to be set as a neighbor of cell A, so the priority of cell B searched by the mobile will be increased too, as well as the success rate of handoff. In this paper, g (NAB/Ntotal) is defined as:

Wk = f (Si )
i =0

Here N is the number of the pilots in the Active Set which votes for the cell K. It obvious that there are two factors may affect the final votes the cell K obtained. One is the strength S of the pilot who vote for cell K. The more the S is, the more the votes will be. Another factor is the number of the pilots which votes for the cell K. The more the N is, the more the votes will be. Sort all the selected cells by their votes obtained descending. Among those, choose all the cells whose votes bigger than zero to make a neighbor cell list. Compared with the algorithm mentioned before, the algorithm presented in this paper not only considers the strength information of pilots in Active Set but also the times that a certain cell selected by summing all the votes it got. It is obvious that the mobile has more probability to move into a cell which selected by several

N ij N AB g N = SLOPE * N + INTERCEPT total total

Here SLOPE and INTERCEPT are parameters stayed to be set by using pilot strength and statistical handoff information. In extreme instance, one pilot is very strong but its NAB/Ntotal is 0, another pilot is very weak but its NAB/Ntotal is 1. It is

reasonable to make this two pilots have equal vote strength, which means

Wi = W j
Considering concluded. and

simultaneously, (7) can be

coefficient dynamic generating algorithm in case of 2 to 4 pilots in the Active Set. The adaptive vote-coefficient dynamic generating algorithm run randomly a million times, select the last ten thousand times to average. Two conclusions can be made form the simulation. 1. The latter two dynamic algorithms take precedence over the simple one (algorithm I). It is obvious that the results got by the latter two dynamic algorithms are not the same when BASE is set to be different value whereas the simple one is independent of the BASE because the pilot strength is not used. BASE is a parameter used to modify the pilot strength tin order to confirm the vote strength of every pilot. It is a little probability that the pilot strength adds BASE more than zero while the BASE is set a quite small value, that is to say the vote strength f (S i) may be a negative or zero. Thus there are few cells that got votes great than zero and the length of final neighbor cell list is short. Along with the increase of BASE, the vote strength f (S i) increases as well. As a result, cells that added into the final neighbor cell list become more and more. It is suitable that BASE is set as 15dB. 2. The result is derived form adaptive vote-coefficient dynamic generating algorithm is more average. The length of the neighbor cell list which resulted from the adaptive vote-coefficient dynamic generating algorithm varies more slowly than the dynamic generating algorithm while the parameter BASE is increasing due to the change of BASE has little effect on the neighbor cell list length considering statistical info. This characteristic is very useful in the actual system.

INTERCEPT * (S i + BASE ) = (SLOP + INTERCEPT ) * (S j + BASE )

In view of unitary requirement, (7) is necessary.


SLOPE + INTERCEPT * NEIGH _ NUM = 1

Finally, SLOPE and INTERCEPT should to be values defined in (9) and (10).
SLOPE = S i + (NEIGHNUM 1)S j + NEIGHNUM * BASE S j + BASE Si S j

INTERCEPT =

S i + (NEIGHNUM 1)S j + NEIGHNUM * BASE

NEIGHNUM in (9) and (10) is the number of neighbor cells of a certain pilot. In former deducibility, NEIGH_NUM of all pilots are equal in default and it may be set as the average of different neighbor cells of all pilots in Active Set. III. SIMULATION AND ANALYSIS

The cells are configured as Figure-1, and then a simulation derived from it can be done. Every sectors neighbor cells are all the sectors bordering upon it and the number will range from 7 to 10.The location of mobile and the pilot strength are set random. The handoff gate T_ADD is -14dB and the value of the pilot strength reported by mobile is -9dB~-16dB.

12 simple insertion dynamic adaptive dynamic vote

10 Length of neighbor-cell list

10

11

12 13 14 BASE (dB)

15

16

17

18

Figure 2.neighbor cell list length at 2 pilots in Active Set

Figure-1 cell arrangement Figure2 to Figure4 show the average of the neighbor cell number respectively built by simple algorithm (algorithm I), the dynamic generating algorithm and the adaptive vote-

12 simple insertion dynamic adaptive dynamic vote

in Active Set a million times. It is obvious that the error rate decreases as the increase of BASE and it is zero when BASE is big than 15dB. So it is reasonable to set BASE as 15dB.
100

10 Length of neighbor-cell list

90 80 70 Error Rate (%)

2 pilots 3 pilots 4 pilots

60 50 40 30

10

11

12 13 14 BASE (dB)

15

16

17

18

20 10 0 10 11 12 13 14 BASE (dB) 15 16 17

Figure 3.neighbor cell list length at 3 pilots in Active Set

14 12 Length of neighbor-cell list 10 8 6 4 2 0 simple insertion dynamic adaptive dynamic vote

Figure 5.Error Rate

IV.

CONCLUSION

10

11

12 13 14 BASE (dB)

15

16

17

18

The suitability of the neighbor cell list has more effect on the success rate of the handoff, thus it is significant to design an effective neighbor cell list generating algorithm. This paper presents a dynamic neighbor cell list generating algorithm based on vote model and optimizes it by using statistical information. By means of simulations it has been shown that the adaptive vote-coefficient dynamic generating algorithm is able to reduce the average of the list without loss of exactitude. That will improve the handoff quality by reducing the mobile search time. The algorithm has been used in a cdma2000 experiment system and performs well. REFERENCES
[1] S.Magnusson and H.OlofssonDynamic neighbor cell list planning in a micro cellular network IEEE 6th International Conference on Universal Personal Communications, Oct 1997, vol. 1, pp 223-227. H.OlofssonDynamic and S.Magnusson A Concept for Dynamic Cell List Planning in a Cellular System 7th IEEE PIMRC , Oct 1996, vol. 1, pp. 138 142. N. Zhang and J. M. Holtzman, "Analysis ofHandoff Algorithms using Both Absolute and Relative Measurements", IEEE VTC 94, 1994,pp. 82-86. J. K. Kwon and D. K. Sung, Soft handoff modeling in CDMA cellular systems, in Proceedingsof the 47th IEEE Vehicular Technology Conference, (Phoenix, AZ), May 1997.

Figure 4.neighbor cell list length at 4 pilots in Active Set The value of BASE also has effect on the error rate of generating neighbor cell list. It is inaccurate to exclude the cell which mobiles often handoff to into the neighbor list owing to little BASE value. In this simulation, if a certain cell A that bordering the cell B, namely cell A and cell B have a public border in figure 1, whose pilot is the highest in Active Set is not included in the final list, an error occurs. Figure 6 shows the result that the adaptive vote-coefficient dynamic generating algorithm respectively runs at 2 to 4 pilots

[2]

[3]

[4]

You might also like