Вы находитесь на странице: 1из 5

A New Call Admission Control Scheme Using the User Conformity for Non-uniform Wireless Systems*

Minhee Chot, Jae-Man Kim$, and Hyunsoo Yoont


+Divisionof Computer Science Department of Electrical Engineering and Computer Science Korea Advanced Institute of Science and Technology E-mail: {mhcho,hyoon} @calab.kaist.ac.kr $Core Network Research Lab. LG Electronics Inc. E-mail : j aeman @ 1ge.com
Absfracf- Next generation wireless communication systems are expected to satisfy quality of service (QoS) requirements by preventing call droppings which are caused by user mobility. Call admission control (CAC) is becoming more important in guaranteeing the QoS of calls. Since existing CAC schemes estimate future requirements and reserve bandwidth to guarantee a certain level of QoS, the bandwidth estimation method is very important in CAC schemes. Most conventional schemes use the movement information of individual mobile terminals (MTs) to estimate the bandwidth requirements. While these schemes estimate the future bandwidth usage relatively well, it requires high computational overhead and lots of memory space to keep and update the movement information of each mobile terminal, and therefore is hard to implement. In this study, we propose a call admission control scheme, which estimates the future handoffs using the handoff histories between BSs and the MT corformify. The proposed scheme achieves low dropping ratio and high bandwidth utilization keeping low computational overhead. The performance of the proposed scheme is evaluated through simulations. Keywords-Call admission control, quality of service, wireless system

I. INTRODUCTION Next generation wireless communication systems are expected to satisfy quality of service (QoS) requirements by preventing call droppings which are caused by user mobility. Call admission control (CAC) is becoming more important in guaranteeing the QoS of calls since more handoffs and more call droppings will occur due to the smaller cell size and increased multimedia applications in future wireless systems. Most CAC schemes try to prevent call droppings by reserving channels which will serve handoff calls, and they reject call requests if channels cannot be reserved. Since existing CAC schemes estimate future requirements and reserve bandwidth to guarantee a certain level of QoS, the bandwidth estimation method is very important in CAC schemes. Most conventional schemes such as the Shadow Cluster [ 11 and the Most Likely Cluster [ 2 ] , etc. use the movement information of individual mobile terminals (MTs) to estimate the bandwidth requirements. While these schemes estimate the future bandwidth us*This work was supported by the Korea Science and Engineering Foundation (KOSEF)through the Advanced Information Technology Research Center (AITrc).

age relatively well, they require high computational overhead and lots of memory space to keep and update the movement information of each mobile terminal, and therefore is hard to implement. Recently, [3] proposed Collective Handoff Probability (CHP)based call admission control strategy, which predicts future channel requests by using only handoff history data among base stations. This strategy assumes that some MTs have similar movement characteristics collectively, due to the objects surrounding the cells such as roads, buildings, or highways. This scheme has the merit that it does not need any complex method to predict each users mobility and therefore has low computational overhead. However, in real environments, while some users show the collectively similar pattern, there exist other users that do not conform to the collective pattern. In the CHP strategy, these non-conforming users cause false handoff prediction, and thus lead to unnecessary call droppings and bandwidth waste. In this study, we propose a call admission control scheme, which estimates the future handoffs using the information of user conformity. A user conformity reflects how strongly a user conforms to the collective handoff probabilities between BSs. Using this information, we can prevent unnecessary bandwidth reservation for non-conforming users and thus get the lower dropping ratio and higher bandwidth utilization. The proposed scheme has low computational overhead because it is simple to keep and update the value of user conformity for each user. The performance of the proposed scheme is evaluated by showing the call dropping ratio and the bandwidth utilization. We compare the performance with two other schemes, the LOCAL and the CHP scheme. In the LOCAL scheme, each base station locally adjusts the amount of reservations needed to achieve the target dropping ratio. The simulation results show that the proposed scheme achieves the lowest dropping ratio and the highest bandwidth utilization for all traffic load ranges. We also examine the effect of different methods for determining the value of user conformity.

0-7803-7005-8/01/$10.00 0 2001 IEEE

1339

This paper is organized as follows. In Section 2, we describe the system model of the mobile system. In Section 3, we provide the simulation results to evaluate the performance of the proposed scheme. The conclusion is presented in Section 4. 11. SYSTEM MODEL A. User Conformity Concept In [ 11, the MT handoff probability was defined as the probability that an M T will leave the current cell to go into an adjacent cell, UJ, given that it enters the cell from a different adjacent cell, U. Although in [ I] it is assumed that the movement information of specific MT, such as the velocity, direction and current location, can easily be obtained, it is known that such information is difficult to obtain in practice [4]. we propose a call admission control scheme, which estimates the future handoffs using the information of user conformity. A user Conformityreflects how strongly an M T conforms to the collective handoff probabilities between BSs. Using this information, we can prevent unnecessary bandwidth reservation for non-conforming MTs and thus get the lower dropping ratio and higher bandwidth utilization. The equations (1)-(3) show our handoff probability estimation, by which we estimate future bandwidth and perform call admission control. Pi,j in equation (1) is defined as the probability that an M T entering the current cell c from the adjacent cell i will leave the current cell to go into the adjacent cell j . P0,j is the CHP for the MTs which originate in the current cell and handoff to cell j . We also define the handoff history, H i , j , as the number of MTs which have entered the current cell from cell i and left the current cell for cell j during a time interval [T,,, - Th, Tcu,].T,,, is the current time and Th is the history collecting duration. The future P,,j is estimated from Hi,j in equation (I).
(3)
f k , the user conformity of MT k , is updated whenever M T IC handoffs We propose two methods for updating the user conformity.

A.0.a User conformity update method I. In method I, f k is updated as equation (4), where fk is the previous value of f k and d is the incremenddecrement for fk as defined in equation ( 5 ) . If the handoff direction of M T k conforms to the collective handoff pattern, its conformity 6 increases by d. Otherwise, its conformity decreases by d. d has different values according to the level of conformity.

fk =

min( f; d , 1) in the conforming case, m a ( & - d, 0) in the non-conforming case

d=

81 if Ri,j 2 el (strongly conform) 6 2 if 1 6 Ri,j < 01 (weakly conform) 82 if 02 6 Ri,j < 1 (weakly unconform) 61 if Ri,j < 02 (strongly unconform)

where is the previous value of Pi,j, N is the number of neighboring cells, and a and p are constants in [0,1]. The user conformity fk and the collective handoff probability Pi,j are used for calculating Q i , j ( k ) in equation (2). The more likely the MT k conforms to the collective handoff pattern, the larger value f k has. P i , j ( k ) , normalized value of the Qi,j ( k ) in equation (3) is the handoff probability actually used for M T k . P , ( k ) represents the probability that an M T k enij tering the current cell c from the adjacent cell i will leave the current cell to go into the adjacent cell j .

The level of conformity is determined using the value of Ri,j in equation (6). 1/N is the probability that an M T entering the current cell from the adjacent cell i will handoff to the adjacent cell j when all the handoffs between cells occur with the same probability. Ri,j represents how high the collective handoff tendency is. Ri,j 2 1represents that the probability of handoff to cell j is larger than the average, while Ri,j < 1 means the tendency to handoffs toward cell j is lower than the average. Assume the M T k entered the current cell c from the adjacent cell i. When it moved to the adjacent cell j where Ri,j 2 1, the MT is said to conform to the collective handoff pattern and f k is increased. On the contrary, f k is decreased when the M T k moved to cell j with Ri,j < 1. Equation (5) shows four conformity levels, where el and 02 (0, > 1 > 02 > 0) are used as thresholds to determine the conformity levels. 81 and 6 2 (61 > 62 > 0) are incremenMdecrernents for strongly and weakly conforming/unconforming case, respectively.

1340

// Pseudo code for new call admission

I 2 3 4 5 6 7

NewCallAdmit(B, z) begin if B BE(t0) > BT then return false for dl t E [to, t o Tmaz] begin if B ;t BE ( t ) BS(t) > BT then If

to-1 Ck=tO-H

Cs(k)+pE(t)

to-1

Ck=tO-H

ci(k)fC:=tO

> PQoS

8 9

then return false end

IO
II
12 13

14 15 16

t i = t o Tmin,t 2 = t o 2Tmax for each cell j E N C begin if CheckNeighborBandwidth(c, t i , t z , B , p0,j (z), 1,z) = false then return false end return true end NewCallAdmit
CheckNeighborBandwidth(c, t l , t 2 , B , p , s t e p , z) begin if BL ( t ) B > BT for any t E [ t l ,t z ] then return false for all t E [ t l ,t z ] begin if B B i ( t ) B: ( t ) > BT then if

// Pseudo code to check the bandwidth of neighboring cells

I
2 3 4

5 6 7
8

C k = t O - H ci(k)+P
to-1

to-1

Ck=tO-H

> c~(k))+c:=to PQoS P;(')+P

9 10 I1 12 13 14

then return false end

B!(t)+ = p . B f o r a l l t E [ t l , t 2 ] = tl Tmin, 4 = tl t 2Trnax if s t e p < M A X S T E P and CheckNeighborBandwidth(j,


t3

t 3 , t4.

B , ~ ! , ~ .(p z s) e p , t

+I

z ) =true for all

k E Nj then

15

return true end CheckNeighborBandwidth Fig. 1. A detailed pseudo code for call admission control

A.0.b User conformity update method 11. In method 1 , fk is 1 updated as equation (7), where f; is the previous value of fk and T is the multiplier for updating f k as defined in equation (8). If the handoff direction of M T k conforms to the collective handoff pattern, its conformity fk increases by T . Otherwise, its conformity decreases by T . T has different values according 1 to the level of conformity as equation (8), where y > 7 2 > 1 > 7 3 > 7 4 and 41 > 1 > 4 2 > 0.

B. CallAdmission Control
When there is a new call request with a bandwidth requirement B in cell e, the proposed scheme entails the following steps: Step 1: Step 1 tests whether admitting this new call will still maintain the QoS of the current cell. Step 2-1: Step 2-1 tests whether admitting this new call will still maintain the QoS of the neighboring cells. N C is the set of neighboring cells of cell c. Then, if the neighboring cells are estimated to be able to maintain their QoS, they reserve bandwidth up to P ; , ~ ( X ) . B, where & ( E ) is the probability that M T z originating in the current cell c will handoff to cell
j.

7 1
T-(

72

73
74

if Ri,j 3 4 1 (strongly conform) if 1 Ri,j < 41 (weakly conform) if 4 2 6 Ri,j < 1 (weaklyunconform) if Ri,j < 4 2 (strongly unconform)

<

(8)

Step 2-2: For each neighboring cell j of cell e, Step 2-2 tests whether the QoS of cell j ' s neighboring cells is satisfied. Step 2-2 is repeated recursively up to the cells which are D,,, distant from cell c. Step 3: If all the cells that have been tested reply that there

1341

will be enough resources, requested bandwidths are reserved in the cells and the new call is admitted to cell c. In Fig. 1, we address a detailed and self-contained pseudo code to describe the overall procedure for call admission control and bandwidth reservation. The notations used in the pseudo code are as follows: t o is the current time. Tmin is the estimated minimum M T dwelling time and T,,, is the estimated maximum M T dwelling time. B l ( t ) is the estimated used bandwidth of cell c at time t , and BF(t) is the reserved bandwidth of cell c at is time t. BT is the total bandwidth of each cell. d,k(x)the probability that an M T 2 entering the current cell c from the adjacent cell i will handoff to the adjacent cell j . H is the time duration for collecting handoff and call dropping history. P Q ~ S is the QoS threshold for call droppings. C i ( k )is the number of handoffs at time k in cell c and C: ( k ) is the number of dropped calls at time k in cell c. d ( k ) is the handoff probability of a call entering cell j at time k .
111. PERFORMANCE EVALUATION

A. Simulation Environment

of each call is exponentially distributed with a mean value of 180s. The dwelling time for each M T is assumed to follow a uniform distribution with the range of 40-60s. Three types of calls are considered, namely, voice, audio or video and the probabilities of these types are 0.7, 0.2 and 0.1, respectively. Each type of call has the following bandwidth requirements: voice = 1, audio = 4, and video = 8 (channels). Each BS has 40 channels. The ratio of the regular to random MT is 3:2. For comparison, we introduce two other schemes, the LOCAL and the CHP schemes. In the LOCAL scheme, each BS locally adjusts the amount of reservations needed to achieve the target drop ratio. Therefore in the LOCAL scheme, a BS increases the reservation amount when the drop ratio is higher than the target drop ratio, and vice versa. The target drop ratio is 0.8% in these simulations. The CHP scheme and the results for regular cells are presented in [3]. The parameters for the CHP scheme and the proposed scheme are as follows: Th = 4s, a = 0.01, ,B = 0.01, Pmin = 0.10, D,,, = 2, Tmin= 40s, and T,,, = 60s. In the conformity update method I of the proposed scheme, O1 = 1.4, 02 = 0.6, 61 = 0.2, and 62 = 0.1. In the conformity update method I1 of the proposed scheme, 41 = 1.4,42 = 0.6, y1 = 1.2, T~ = 1.1,y3 = 0.9, and 7 4 = 0.8.
C. Simulation Results

The simulation model is composed of 5 x 5 cells with wraparound edge cells. Two types of MTs are assumed: a random MT, which has the same handoff probabilities for all neighboring cells, and a regular MT, which hands off only to some particular cells. A cell having only random MTs is defined as a normal cell and a cell having both random and regular MTs as a regular cell. In Fig. 2, the cells on the horizontal and vertical road are regular cells, and other cells are normal cells. Regular cells have a greater traffic load and more handoff attempts than normal cells because they deal with both normal and regular MTs.

Fig. 3 shows the performance of dropping ratio against the offered load for the schemes. The proposed scheme achieves the lowest dropping ratio for all traffic load ranges. The dropping ratio for the CHP and the proposed schemes decreases from a certain traffic load, because, as the traffic load increases, the attempts of the CAC schemes to reserve bandwidth in neighboring cells fail more often, and traffic with larger bandwidth requirements is more often rejected. Therefore, because more bandwidth can be saved for handoff calls, the dropping
ratio decreases.

0.014

0.012

0.01

0.00R
0.006
0.004

s
Fig. 2. Simulation Environment

0.002
0 '
" " " "

"
I

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.x 0.9

0 tcrcd load 1

B. Simulation Parameters
We assume that new call requests follow a Poisson distribution with variable mean arrival rates and that the holding time

Fig. 3. Dropping probability vs. offered load

Fig. 4 shows the bandwidth utilization performance for the

1342

schemes. The proposed scheme achieves a higher bandwidth utilization than the LOCAL scheme and the CHP scheme. It is because the proposed scheme estimates the handoff probability of MT more accurately with small overhead.
3

5
0.7 0.6 0.5 0.4 0.3 0.2

0.45 04 .

B
0.35
0.3

Proposed B -+---.

-.

Fig. 6. Bandwidth utilization vs. offered load

0 1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 9 . .


Offered load

Fig. 4. Bandwidth utilization vs. offered load

Fig. 5 shows the dropping ratio of the proposed A and the proposed B schemes. The proposed A and the proposed B schemes update the user conformity using method I and method 11, respectively. The proposed B scheme achieves lower dropping ratio than the proposed A scheme. Fig. 6 shows the bandwidth utilization of the proposed A and the proposed B scheme. The proposed B scheme provides a little higher bandwidth utilization than the proposed A scheme. Hence, we can see that the update method I1 is more effective than the update method I.
0.014 0.012 00 .1
Proposedl B -*---

reflects how strongly a user conforms to the collective handoff probabilities between BSs. Using this information, we can prevent unnecessary bandwidth reservation for non-conforming users and thus get the lower dropping ratio and higher bandwidth utilization. The proposed scheme has low computational overhead because it is simple to keep and update the value of user conformity for each user. The performance of the proposed scheme is evaluated by showing the call dropping ratio and the bandwidth utilization. The simulation results show that the proposed scheme achieves the lowest dropping ratio and the highest bandwidth utilization for all traffic load ranges.

REFERENCES
D. A. Levine, I. F. Akyildiz, and M. Naghshineh, A Resource Estimation and Call Admission Algorithm for Wireless Multimedia Networks Using the Shadow Cluster Concept, IEEE Trunsactions on Networking, vol. 5, no. I , pp. 1-12, Feb. 1997. A. Aljadhai and T. F. Znati, A Framework for Call Admission Control and QoS Support in Wireless Environments, in Pruc. IEEE INFOCOM99, 1999, pp. 1019-1026. Jae-Man Kim, Eui-Hoon Jeong and Jung-Wan Cho, Call Admission Control for Non-uniform Traffic in Wireless Networks, Electronics Letters, vol. 36, no. 1, pp. 96-97, Jan. 2000. C. Oliveira, J. B. Kim, and T. Suda, An Adaptive Bandwidth Reservatoin Scheme for High-speed Multimedia Wireless Networks, IEEE SAC, vol. 16, no. 6, pp. 858-874, August 1998.

0 0

1
1

0 1 0.2 .

0.3

0.4 0.5 0.6 Offercd load

0.7 0.8 0.9

Fig. 5. Dropping probability vs. offered load

IV. CONCLUSION We have proposed a new CAC scheme for wireless network systems. The proposed scheme utilizes handoff information between base stations and user conformity. A user Conformity

1343

Вам также может понравиться