J. Cent. South Univ. Technol. (2010) 17: 110-116

DOI: 10.1007/s11771-010-0018-2                                                                                                                

A novel dynamic call admission control policy for wireless network

HUANG Guo-sheng(黄国盛)^{1, 2}, CHEN Zhi-gang(陈志刚)¹, LI Qing-hua(李庆华)¹,

ZHAO Ming(赵明)¹, GUO Zhen(郭真)¹

1. School of Information Science and Engineering, Central South University, Changsha 410083, China;

2. College of Physics Science and Information Engineering, Jishou University, Jishou 416000, China

? Central South University Press and Springer-Verlag Berlin Heidelberg 2010

Abstract: To address the issue of resource scarcity in wireless communication, a novel dynamic call admission control scheme for wireless mobile network was proposed. The scheme established a reward computing model of call admission of wireless cell based on Markov decision process, dynamically optimized call admission process according to the principle of maximizing the average system rewards. Extensive simulations were conducted to examine the performance of the model by comparing with other policies in terms of new call blocking probability, handoff call dropping probability and resource utilization rate. Experimental results show that the proposed scheme can achieve better adaptability to changes in traffic conditions than existing protocols. Under high call traffic load, handoff call dropping probability and new call blocking probability can be reduced by about 8%, and resource utilization rate can be improved by 2%-6%. The proposed scheme can achieve high source utilization rate of about 85%.

Key words: wireless network; call admission control; quality of service; Markov decision process

1 Introduction

With the rapid development of wireless network communication, effective use of the limited wireless resources and QoS support in wireless network has become more and more important, and considerable efforts have been focused on call admission control (CAC). The lack of wireless resources is the major bottleneck in wireless networks to provide QoS support. The next generation of wireless cellular networks will use micro/pico cellular architectures in order to provide higher capacity, mobile nodes handovering frequently among cells brings new challenges to the call admission control [1-3]. An ongoing call in the current cell may have to be handed over to another cell. During this process, the call may not be able to obtain enough resources to continue its communication due to the limited available resources in the new cell, which will lead to call dropping. Because mobile users are more sensitive to ongoing call dropping than new call blocking, handoff calls are normally assigned higher priority over new calls [4-5].

There are two important connection level QoS parameters in wireless communication: new call blocking probability P_nb and handoff call dropping probability P_hd. CAC optimization goal is to maximize resource utilization while reduce P_hd and P_nb. However, in order to reduce P_hd, the existing CAC policies usually result in obvious decrease of resource utilization and may lead to a high P_nb [6].

The existing CAC schemes can be divided into two major categories: reservation-based schemes and threshold-based schemes [7-8].

In Ref.[9], when a handoff occurred, bandwidth was allocated in the new cell and reserved in the new cell’s neighboring cells, according to the call connection number in neighboring cells and the mobility direction of the mobile node. The disadvantage of this scheme is that the reserved bandwidth in the neighboring cells would cause a waste of system resources, resulting in the increase of P_nb in these cells. GWENDAL and ERIC [10] used mobile IP reservation protocol (MIR) to reserve a fixed bandwidth in each cell for switching users, carrying out CAC according to the reserved bandwidth-size. However, MIR solution did not give an algorithm to determine the reserved bandwidth-size. It is difficult for this solution to adapt the dynamic changing of network conditions.

Guard channel (GC) solution [11-13], is the most widely used threshold-based CAC solution. New call bounding (NCB) scheme, proposed by FANG and ZHANG [13], is a typical representative of GC solutions. The scheme works as follows: when a new call arrives, if the number of new calls in a cell exceeds a threshold K, the new call will be blocked; otherwise it will be accepted. The handoff call will be rejected only when all channels in the cell are used up. In fact, GC solutions are equivalent to specifically reserve part of system bandwidth for handoff calls. In threshold-based solution, P_hd is reduced at cost of increase of P_nb. If the reserved bandwidth for handoff calls is too large, although P_hd is reduced, a rapid increase of P_nb will cause a decline in integral performance of system connection level QoS. Furthermore, part of GCs may be left to be idle, resulting in the decrease of system resource utilization. On the contrary, if reserved bandwidth for handoff calls is too small, the P_nb will come down while P_hd will go up, which is difficult to meet the QoS demand of mobile users [14-17].

WANG et al [18] proposed an adaptive call admission control solution. The solution calculated the optimal call number of each cell adaptively according to the system state vector, call end vector and call transfer matrix. It performed CAC based on NCB scheme. The advantage of this solution is that it can adapt to the changes of network load. However, this solution needs to calculate the system call transfer matrix and overload probability to find the ideal system state [18]. With the increase of the cell number of a domain, the computational load will increase considerably. Although some adaptive GC based CAC solutions have been proposed, it is difficult for existing schemes to reduce P_hd and P_nb effectively and improve system resources utilization [19-20].

In this work, a reward mechanism based dynamic optimization scheme on CAC (RBDO-CAC) for wireless mobile network was proposed. The scheme established a reward model of call admission control of wireless cell based on Markov decision process, dynamically optimized call admission process according to the principle of maximizing the average system rewards, and had a good adaptability to changes in network flow conditions.

2 System model analysis

RBDO-CAC was based on NCB scheme [13]. It adapted to the changes of system call traffic load, adjusted new call admission threshold K of a cell according to the principle of maximizing system reward, consequently, realized the dynamic optimization of CAC.

As shown in Fig.1, calls in each wireless cell can be divided into two types: new calls and handoff calls. New call is initiated from mobile nodes (MNs) of current cell. Handoff call is an ongoing call moving from adjacent cell to current cell. Assume the effective capacity of    a cell is C, then C represents the maximum connection requests that the cell can accept [21-22].

Fig.1 Structure model of wireless cell

To simplify the problem description, a cell was chosen as the research object. Assume arrival process of new calls and handoff calls follow the Poisson distribution, and their arrival rates are λ₁ and λ₂ respectively. Service time of new calls and handoff calls follows the exponential distribution, and their average service time is 1/μ₁ and 1/μ₂, respectively. The cell channel occupied state can be expressed by a two- dimensional Markov chain, and the system state space can be written as follows:

S={(n₁, n₂)│0≤n₁≤K, n₂≥0, n₁+n₂≤C}          (1)

where n₁ denotes the number of new calls accepted in the cell, and n₂ is the number of handoff calls in the cell.

For the purpose of simplicity, the following definitions are given.

Definition 1  New call traffic intensity ρ₁ is the multiplication of new call arrival rate and new call average service time, i.e., ρ₁=λ₁/μ₁.

Definition 2 Handoff call traffic intensity ρ₂ is the multiplication of handoff call arrival rate and handoff call average service time, i.e., ρ₂=λ₂/μ₂.

According to the above definitions, the larger the ρ₁ and ρ₂ are, the heavier the system load is.

Assume the system rewards of accepting a new call and handoff call are R_n and R_h, respectively. When the system is in state (n₁, n₂), the total reward R_T can be written as follows:

R_T=n₁R_n+ n₂R_h                                                 (2)

In RBDD-CAC, the new call admission threshold K was dynamically optimized to maximize the system average reward . Thus, it can be known that

 K∈[1, C]        (3)

where K_m is the optimal new call threshold to maximize the system average reward. Obviously, to give priority to handoff calls, the reward of accepting a new call and a handoff call (i.e., R_n and R_h) must satisfy:

R_h＞R_n                                                         (4)

The larger the ratio of R_h to R_n, the higher the priority P of handoff calls, that is

P ∝ R_h/R_n                                   (5)

The system state transition diagram is shown in Fig.2, where K stands for new call threshold, C stands for system capacity.

Fig.2 Transition diagram of system state

The schematic diagram of system state transition rate between adjacent states is shown in Fig.3, where   (i, j) stands for a state in state space S, i denotes the number of new calls initiated in the cell and j is the number of handoff calls in the cell, i∈[1, K-1] and j∈[1, C-1].

Fig.3 Schematic diagram of system state transition rate

Assume P(n₁, n₂) denotes the steady state probability of n₁ new calls and n₂ handoff calls in the cell, then the system state balance equation can be written as

λ₁P(i-1, j)+(i+1)μ₁P(i+1, j)=iμ₁P(i, j)+λ₁P(i, j),

i∈[1, K-1], j∈[1, C-2]                    (6)

λ₂P(i, j-1)+(j+1)μ₂P(i, j+1)=jμ₂P(i, j)+λ₂P(i, j),

i∈[1, K-1], j∈[1, C-2]                    (7)

λ₁P(i, 0)=μ₁P(1, j), j∈[0, C-1]                  (8)

λ₂P(i, 0)=μ₂P(i,1), i∈[0, K]                     (9)

Cμ₂P(0, C)=λ₂P(0, C-1)                       (10)

From Eqs.(6)-(10), it can be known that

λ₁P(i , j)=(i+1)μ₁P(i+1, j),  0≤i≤K-1, 0≤j≤C   (11)

λ₂P(i , j)=(j+1)μ₂P(i, j+1),  0≤i≤K, 0≤j≤C-1   (12)

According to system state balance equation, by using recursive method, the steady state probability P(n₁,n₂) can be expressed as [13]

0≤n₁≤K, n₁+n₂≤C (13)

Then, the normalization equation can be written as

                         (14)

According to Eqs.(13) and (14), it can be known that

              (15)

According to Eqs.(13) and (15), new call blocking probability P_nb and handoff call dropping probability P_hd in the cell can be written as

          (16)

    (17)

According to Eq.(13), the mathematical expectation of new call number and handoff call number in the cell can be expressed as

           (18)

          (19)

According to Eqs.(2), (18) and (19), the system average reward  of the cell can be expressed as

=

          (20)

According to Eqs.(3) and (20), it is not difficult to find the optimal new call threshold K_m to maximize system average reward.

3 Reward mechanism based dynamic optimization on call admission control

In RBDO-CAC, the optimal call admission threshold K_m was calculated, according to the model established in section 2 based on the principle of maximizing system average reward, and K_m was dynamically optimized in time period τ in accordance with the real-time call traffic load.

Because the system obtains more reward accepting a handoff call than accepting a new call, so the system will give priority to accept handoff calls, that is, handoff call has a higher priority than new call. At the same time, in order to obtain the largest average reward, the system will make full use of available network resources to accept new calls as many as possible. Consequently, RBDO-CAC can improve system resource utilization and optimize the system performance.

When a call request arrives, the system first checks the call type, for new calls, only when the new call number system accepted is less than threshold K_m, it can be accepted; for handoff calls, as long as the system has available resources, it will be accepted.

According to the call admission model given in section 2, reward mechanism based dynamic optimization on call admission control algorithm is described as follows:

// Pseudo code of dynamic optimization CAC

// Input:

// C= effective capacity of a cell;

// τ = time period for updating threshold K_m;

// λ₁ = arrival rate for new call;

// λ₂ = arrival rate for handoff call;

// 1/μ₁= average channel holding time for new call;

// 1/μ₂ = average channel holding time for handoff call;

// Temporary variables:

// R_max: a temporary variable to store system reward;

// K₁: a temporary variable to store call threshold;

// K_m: the optimal call admission threshold for new call;

// C_A: number of accepted new call in the cell;

// C_B: total number of accepted calls in the cell;

for ( ; ; )

// Find the optimal call admission threshold K_m

{ for (j= 1; j≤C ; j++)

{k = j;

compute  according to Eq.(20);

if (R_max＜) {R_max=; K₁=k}

}

K_m=K_1;

// K_m is the optimal call admission threshold

// Call admission control with threshold K_m

if (Connection type = = new call)

{if (C_A+1＜K_m and C_B＜C)

{accept; C_A = C_A+1; C_B = C_B +1}

else  reject; }

else  // Connection type is handoff call

{if (C_B+1＜C)

{accept; C_B = C_B+1;}

else  reject; }

Delay time-interval τ;

//Update threshold K_m periodically

}

In this algorithm, the optimal new call admission threshold K_m is calculated dynamically. The algorithm can maximize system reward, improve system resource utilization, and adapt to the dynamic changes of network traffic.

4 Simulation and performance analysis

RBDO-CAC was simulated in Red Hat Linux. The effective capacity of a cell is 50. To simplify the calculation, the system reward of accepting a new call is given as R_n=1, by changing R_h to adjust the priority of handoff calls. The optimal new call admission threshold K_m, new call blocking probability P_nb, handoff call dropping probability P_hd and system resource utilization R_u under different call traffic intensity (ρ₁and ρ₂) and different handoff call reward R_h, were measured, and the simulation results were compared with NCB solution [13].

4.1 Optimal new call admission threshold K_m

Fig.4 shows the relationship between system average reward and new call admission threshold K when maintaining handoff call reward R_h=3 and new call traffic intensity ρ₁=35 unchanged, and it also shows the effect of changing handoff call traffic intensity ρ₂ on the optimal threshold K_m. It can be seen from Fig.4 that when handoff call traffic intensity ρ₂ is 22, 24 and 26, the new call admission threshold K_m of obtaining the maximum system average reward is 26, 22 and 20, respectively, which indicates that the greater the ρ₂, the smaller the  K_m. This is because the greater the ρ₂, the more the handoff call admission request. The system needs to reduce new call threshold K_m to accept more handoff calls to obtain the largest reward.

Fig.4 Optimal call threshold K_m under different handoff call traffic intensities ρ₂

Fig.5 shows the relationship between system average reward and new call admission threshold when maintaining ρ₁=35, ρ₂=25 unchanged. As shown in Fig.4, when handoff call reward R_h=2.5, 3.0 and 3.5, the system optimal call threshold K_m=23, 21 and 20, respectively. It is found that the greater the R_h, the smaller the K_m. This is because, according to the principle of maximizing system reward, the greater the R_h, the higher the priority of handoff calls, the system will reduce K_m to accept more handoff calls.

Fig.5 Optimal call threshold K_m under different handoff call reward R_h

4.2 Handoff call dropping probability P_hd and new call blocking probability P_nb

Fig.6 shows handoff call dropping probability of RBDO-CAC and NCB when new call traffic intensity ρ₁=20, and handoff call traffic intensity ρ₂ changes from 21 to 35. As can be seen from Fig.6, with the increase of handoff call traffic intensity ρ₂, handoff call dropping probabilities P_hd of the two solutions both increase, but P_hd of RBDO-CAC has a more gentle change. It is found that RBDO-CAC has a better adaptability to network call traffic changes, and P_hd of RBDO-CAC is lower than that of NCB. This enhancement is expected as ρ₂ increases, DT-CAC adaptively reduces new call admission threshold K_m, accepting more handoff calls to maximize system reward, consequently, limiting the increase of P_hd. In addition, as shown in Fig.6, for RBDO-CAC, the larger the handoff calls reward R_h, the smaller the P_hd. This is because the greater the R_h, the higher the priority of handoff call, and the more the handoff call accepted.

Fig.6 Handoff call dropping probability P_hd vs handoff call traffic intensity ρ₂

Fig.7 shows the relationship between P_nb and ρ₁ of RBDO-CAC and NCB when handoff call traffic intensity ρ₂=21. As shown in Fig.7, for RBDO-CAC, the greater the handoff call reward R_h, the greater the P_nb. This is because in accordance with the principle of maximizing system average reward, the greater the R_h, the higher the priority of handoff calls, consequently, the larger the P_nb. Instead, the smaller the handoff call reward R_h, the smaller the P_nb. So a good trade-off between P_hd and P_nb can be achieved by adjusting handoff call reward. From Fig.7, it can also be seen that probabilities P_nb of two solutions both increase when ρ₁ increases, the increase of P_nb of NCB is more significant, and P_nb of RBDO-CAC is less than P_nb of NCB. This is because in NCB, when new call traffic intensity ρ₁ increases, call admission thresholds remain unchanged. RBDO-CAC can adapt to changes of network load and dynamically optimize call admission threshold according to the principles of maximizing system average reward, accepting more new calls, thus restricting the increase of P_nb.

Fig.7 New call blocking probability P_nb vs new call traffic intensity ρ₁

4.3 Resource utilization under different call traffic intensities

Fig.8 depicts system resource utilization of RBDO-CAC and NCB under different new call traffic intensity ρ₁, where handoff call traffic intensity is given as ρ₂=20, and handoff call reward of RBDO-CAC is given as R_h=2.0. As shown in Fig.8, when ρ₁ increases, resource utilizations of the two solutions both increase, and resource utilization of RBDO-CAC is significantly higher than that of NCB. Because when ρ₁ increases, call admission threshold of NCB remains unchanged, which cannot adapt to the changes of network load. On the contrary, RBDO-CAC can dynamically optimize new call admission threshold K_m, take full advantage of available channels in the cell to accept more calls, maximize system average reward, and thus improve system resources utilization.

Fig.8 System resource utilization of RBDO-CAC and NCB under different new call traffic intensities ρ₁

5 Conclusions

(1) The call admission control scheme proposed dynamically calculates call admission threshold according to network load conditions, and can adapt to the dynamic changes of call traffic.

(2) Realization of maximizing system rewards enables system to accept calls request as many as possible, and effectively improves system resources utilization.

(3) The scheme effectively reduces system handoff call dropping probability and new call blocking probability, achieves a good trade-off between handoff call dropping probability and new call blocking probability by adjusting the reward of handoff call and new call, and prevents unrestricted increase of new call blocking probability caused by the priority of handoff call.

References

[1] DIMITRIOS G, GEORGIOS I, PPANAYOTIS G. Call admission control in wireless networks: Probabilistic approach and efficiency evaluation [C]// Proceedings of Wireless Communications and Mobile Computing Conference. Crete Island: IEEE, 2008: 712-717.

[2] MA Yu-feng, GONG Shen-guang, HU Xiu-lin, ZHANG Yun-yu. Analysis on call admission control in cellular wireless communication networks [J]. Journal on Communications, 2006, 27(5): 107-113. (in Chinese)

[3] NASSER N, HASSANEIN H. An optimal and fair call admission control policy for seamless handoff in multimedia wireless networks with QoS guarantees [C]// Proceedings of the IEEE Global Telecommunications Conference(GLOBECOM’04). Kingston: IEEE Communications Society, 2004: 3926-3930.

[4] MEHRDAD M, HAMIDREZA B, MOSTAFA P. A new dynamic pricing scheme with call admission control to reduce network congestion [C]// Proceedings of 22nd International Conference on Advanced Information Networking and Applications. Okinawa: IEEE, 2008: 348-352.

[5] HUANG Guo-sheng, CHEN Zhi-gang, ZHAO Ming, QI Hua-mei, LIANG Ping-yuan, HU Yu-ping. Link layer assisted end-to-end QoS guarantee scheme in hierarchical mobile IPv6 [J]. Journal of System Simulation, 2008, 20(21): 5852-5862. (in Chinese)

[6] WANG Sheng-ling, HOU Yi-bin, HUANG Jian-hui, HUANG Zhang-qin. Adaptive call admission control based on enhanced genetic algorithm in wireless/mobile network [C]// Proceedings of the 18th IEEE International Conference on Tools with Artificial Intelligence (ICTAI'06). Washington: IEEE Communications Society, 2006: 3-9.

[7] CHEN H, CHENG C, YEH H. Guard channel based incremental and dynamic optimization on call admission control for next-generation QoS-aware heterogeneous systems [J]. IEEE Transactions on Vehicular Technology, 2008, 57(5): 3064-3082.

[8] ZHANG Xue. Survey on call admission control models in wireless mobile networks [J]. Journal on Communications, 2005, 26(8): 99-106. (in Chinese)

[9] OLIVIERA C, KIM J B, SUDA T. An adaptive bandwidth reservation scheme for high speed multimedia wireless networks [J]. IEEE Journal of Selected Areas in Communications, 1998, 16(6): 858-874.

[10] GWENDAL L, ERIC H. A predictive end-to-end QoS scheme in a mobile environment [C]// Proceedings of the Sixth IEEE Symposium on Computers and Communications. Hammamet, Tunisia: IEEE, 2001: 534-539.

[11] SHIOKAWA S, ISHIZAKA M. Call admission scheme based on estimation of call dropping probability in wireless networks [C]// Proceedings of the 13th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications. Lisboa: IEEE, 2002: 2185-2188.

[12] HONG D, RAPPORT S S. Traffic model and performance analysis for cellular mobile radio telephone systems with prioritized and nonprioritized handoff procedures [J]. IEEE Transactions on Vehicular Technology, 1986, 35(3): 77-92.

[13] FANG Y, ZHANG Y. Call admission control schemes and performance analysis in wireless mobile networks [J]. IEEE Transactions on Vehicular Technology, 2002, 51(2): 371-382.

[14] JIANG Ai-quan, ZHAO A-qun. Adaptive admission control algorithm and performance analysis in wireless/mobile networks [J]. Journal on Communications, 2004, 25(6): 147-156. (in Chinese)

[15] CONG Qiu-shi, JIANG Ai-quan, PAN Ji-ming. Admission control scheme using pricing incentive in wireless/mobile networks [J]. Computer Engineering and Applications, 2006, 42(33): 125-128. (in Chinese)

[16] FANG Xu-ming, ZHANG Dan-dan. Overview on call admission control schemes in wireless communication networks [J]. Computer Applications, 2006, 26(8):1762-1767. (in Chinese)

[17] XU Xing, YE Wu, FENG Sui-li. A novel calling admission control scheme based on multi-priority delay reservation [J]. Journal of Beijing University of Posts and Telecommunications, 2007, 30(1): 75-79. (in Chinese)

[18] WANG Sheng-ling, HOU Yi-bin, HUANG Jian-hui, HUANG Zhang-qin. Hierarchical mobile IPv6 based on adaptive threshold call admission control [J]. Journal of Software, 2006, 17(9): 1996-2003. (in Chinese)

[19] HUANG Guo-sheng, CHEN Zhi-gang, ZHAO Ming, WANG Lu-lu. New multicast scheme based on dynamic mobility prediction in mobile IPv6 environment [J]. Journal of Central South University of Technology, 2007, 14(3): 386-392.

[20] YAVUZ E A, LEUNG V C. Methods for performance evaluations of call admission control schemes in multi-service cellular networks [J]. IEEE Transactions on Wireless Communications, 2008, 7(9): 3468-3476.

[21] OH E, HAN S, WOO C, HONG D. Call admission control strategy for system throughput maximization considering both call- and packet-level QoSs [J]. IEEE Transactions on Communications, 2008, 56(10): 1591-1594.

[22] CHATZIPERIS S, KOUTSAKIS P, PATERAKIS M. A new call admission control mechanism for multimedia traffic over next-generation wireless cellular networks [J]. IEEE Transactions on Mobile Computer, 2008, 7(1): 95-111.

Foundation item: Project(60873082) supported by the National Natural Science Foundation of China; Project(09C794) supported by the Natural Science Foundation of Education Department of Hunan Province, China; Project (S2008FJ3078) supported by the Science and Technology Program Foundation of Hunan Province, China; Project(07JJ6109) supported by the Natural Science Foundation of Hunan Province, China

Received date: 2009-02-28; Accepted date: 2009-07-09

Corresponding author: CHEN Zhi-gang, Professor; Tel: +86-731-88830797; E-mail: czg@mail.csu.edu cn

(Edited by CHEN Wei-ping)