Você está na página 1de 5

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the IEEE Globecom 2010 proceedings.

Mobility-Aware Call Admission Control Algorithm in Vehicular WiFi Networks


Younghyun Kim, Sangheon Pack
School of Electrical Engineering Korea University, Korea Email: {younghyun kim, shpack}@korea.ac.kr

Wonjun Lee
Department of Computer Science and Engineering Korea University, Korea Email: wlee@korea.ac.kr

AbstractResource management in vehicular WiFi networks is an interesting and challenging issue. In this paper, we propose a mobility-aware call admission control (MA-CAC) algorithm where different admission control policies are employed depending on the mobility. Specically, when a vehicle is static, a handoff priority scheme with guard channels is examined to protect vehicular handoff users. On the other hand, for a moving vehicle, no guard channels for handoff users are allocated for maximizing channel utilization since there are no vehicular handoff users. By means of Markov chains, we evaluate the MACAC algorithm in terms of new call blocking probability, handoff call dropping probability, and channel utilization. Numerical results demonstrate that the MA-CAC algorithm can lower the handoff call dropping probability while maintaining high channel utilization.

I. I NTRODUCTION Currently, high-speed rail systems in Korea (a.k.a. Korea Train eXpress (KTX)) provide in-vehicle wireless Internet services by means of internal wireless local area network (WLAN) connections and external high speed downlink packet access (HSDPA) connections. In addition, portable mobile relays, which is a gateway between WLANs and HSDPA or Mobile WiMAX (a.k.a. WiBro in Korea), allow Internet access via WLAN interfaces, and they are available in the market (e.g., Egg from Korea Telecom (KT)). Moreover, Seoul city starts free Internet services in taxi using these mobile relays. Since users spend more and more time in vehicles (e.g., car, bus, subway, and train), these vehicular WiFi networks will be more popular in the future. In our previous work [1], unique features (i.e., heterogeneous wireless links with WLAN and wireless wide area network (WWAN)) in vehicular WiFi networks are identied and analyzed. In addition, the results in [1], [2] reveal that an efcient resource management is indispensable. Moreover, our measurement study in KTX and car [3] conrms that current Internet services in vehicular WiFi networks are not satisfactory due to the absence of resource management schemes. In particular, uplink throughput is unmatched with the requirements of interactive multimedia applications such as voice or multimedia over IP. In this work, we focus on call admission control (CAC) issues at the access point (AP) installed on the vehicle. Since the WLAN capacity of the AP is shared by multiple vehicular users, a well-dened CAC algorithm is needed to

satisfy quality of service (QoS) requirements. Since CAC is a fundamental issue for QoS provision [4], extensive works have been done in the literature (see [5] for a comprehensive survey). Recent works consider some challenging issues such as multi-class services and vertical handover. For example, Liao et al. [6] proposed a closed-loop feedback control system for multi-class services, which integrates priority-based queueing model, adaptive fuzzy service degradation control model, and optimal service degradation location model. On the other hand, Lee et al. [7] proposed a probabilistic call admission control algorithm for reducing unnecessary vertical handover while maintaining high resource utilization in WLAN area. In terms of vehicular WiFi networks, vehicular user mobility (i.e., riding on or getting off the vehicle) should be taken into account for designing efcient CAC algorithms. However, to the best of our knowledge, no works on CAC in vehicular WiFi networks are reported. In this paper, we propose a mobility-aware call admission control (MA-CAC) algorithm exploiting unique features in vehicular WiFi networks. The MA-CAC algorithm introduces two phases: stop and moving. At the stop phase, a vehicle stays at a location (e.g., bus stop or subway station) and thus vehicular users can join to the AP (i.e., riding on the vehicle). In this situation, a handoff priority scheme with guard channels is examined to protect vehicular handoff users1 from vehicular new users. On the other hand, at the moving phase, there are no vehicular handoff users since no users can ride on the moving vehicle. Hence, no guard channels for handoff users are allocated to maximally utilize the AP resource. By means of Markov chains, we evaluate the MACAC algorithm in terms of new call blocking probability, handoff call dropping probability, and channel utilization. Numerical results demonstrate that the MA-CAC algorithm can lower the handoff call dropping probability under high channel utilization. The remainder of this paper is organized as follows. Section II presents the system model. Section III describes the MA-CAC algorithm. Analytical model based on Markov chains and numerical results are given in Sections IV and V, respectively. Finally, Section VI conclude this paper.
1 Vehicular handoff users refer to ones who have made calls with an external base stations (BS) and perform a handoff from external BS to the AP at the vehicle [8].

978-1-4244-5638-3/10/$26.00 2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.

Fig. 1.

Vehicular WiFi networks. Fig. 2. Mobility-aware call admission control scheme.

II. S YSTEM M ODEL Consider a scenario where a vehicular user with an ongoing call rides on a vehicle, i.e., mobile node (MN)1 in Figure 1. Then, the MN1 should perform a handoff from the BS to the AP installed in the vehicle and thus the call is referred to as a handoff call. On the other hand, the vehicular user can initiate a new call after riding on the vehicle and then the call is considered as a new call with respect to the AP (e.g., MN2 in Figure 1). In vehicular WiFi networks, there are two types of handoff events: one is from the AP to the BS and another is from the BS to the AP. For a former handoff, general CAC algorithms will be performed at the BS. On the other hand, for a latter case, a new CAC algorithm is needed to consider vehicular mobility, which is the motivation of this work. Therefore, hereinafter, we consider only the handoff from the BS to the AP or CAC at the AP. CAC algorithms achieve efcient resource management by limiting the number of admitted calls and ensuring QoS to users. Since mobile users experience severe service disruption when handoff calls are dropped, many CAC algorithms employ guard channels for prioritizing and protecting handoff calls. The proposed MA-CAC algorithm adopts the concept of guard channels for handoff calls from the BS to the AP. However, unlike the existing CAC algorithms, the guard channels are dynamically assigned by exploiting the vehicular mobility. That is, some guard channels are allocated for handoff calls when the vehicle is static (e.g., at the stop) and vehicular users are riding on. On the other hand, since handoff calls cannot be generated during the movement, the remaining WLAN channels can be used by new calls without any guard channels. Detailed algorithm will be described in Section III. To describe the MA-CAC algorithm, we consider the discrete WLAN capacity as in [9], i.e., the WLAN capacity is denoted by C bandwidth units. This WLAN capacity C can be determined by considering service requirements, e.g., VoIP capacity can be obtained as [10]. In addition, we assume that each call requires a single bandwidth unit. III. M OBILITY-AWARE C ALL A DMISSION C ONTROL The MA-CAC algorithm denes two phases: stop and moving. At the stop phase, the vehicular users can ride on or get off the vehicle. In other words, handoff calls can be generated at the stop phase. Therefore, the MA-CAC algorithm sets guard channels for ensuring higher priorities to handoff calls. Specically, the MA-CAC algorithm employs the cutoff priority (CP) scheme in which a xed number of channels (i.e., C K channels) are exclusively used for handoff calls [11]. However, note that the MA-CAC algorithm can utilize other handoff priority-based schemes. After stopping at a location, the vehicle moves to another location and then the phase is changed from the stop phase to the moving phase. Since no vehicular users cannot ride on the vehicle during the moving phase, we do not consider any handoff calls for call admission control. In other words, it is wasteful to reserve a portion of channel (i.e., guard channels) for handoff calls during the moving phase. Consequently, the total WLAN capacity C can be used by new calls at the moving phase. Figure 2 shows the ow chart of the MA-CAC algorithm. When a call arrives, the AP checks whether the call is a handoff call or a new call. In the case of handoff call, the call is accepted if the number of total calls is less than C. On the contrary, when a new call arrives, the phase should be examined. For the stop phase, if the number of occupied channels is less than K, the new call is admitted; otherwise, it will be blocked. In the moving phase, the new call is rejected only when the number of total calls is equal to C. Recent mobile terminals have multiple interfaces such as WLAN and cellular systems. The blocked call can be redirected to cellular networks in WLAN/cellular interworking systems. This is another interesting topic in vehicular WiFi networks; but it is out of scope of this paper. IV. P ERFORMANCE A NALYSIS In this section, we evaluate the performance of the MA-CAC algorithm using continuous time Markov chains (CTMCs), in terms of the new call blocking probability (NCBP), handoff call dropping probability (HCDP), and channel utilization. To develop CTMCs, we have the following assumptions: The arrival processes of new calls and handoff calls follow Poisson distributions with rates N and H ,

978-1-4244-5638-3/10/$26.00 2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.

Fig. 3.

Two-dimensional CTMC for the stop phase.

respectively. The call duration times of new calls and handoff calls follow exponential distributions with means 1 and 1 , N H respectively. Depending on the phase, the MA-CAC algorithm requires different CTMCs and thus we analyze the stop and moving phases, separately.

Fig. 4.

Two-dimensional CTMC for the moving phase.

A. Stop Phase In the stop phase, the CP scheme is employed for handoff call prioritization. Figure 3 shows the state transition diagram for the stop phase, where a state (i, j) represents that i new calls and j handoff calls. Then, the state space is given by S = {(i, j)|0 i K, 0 i + j C}, where K and C are the pre-dened threshold and total capacity, respectively. Let p(i, j; denote the transition rate from state (i, j) to state i, j) ( Then, the transition rates are summarized as i, j). p(i, j; i 1, j) = iN (0 < i K, 0 i + j C) p(i, j; i + 1, j) = N (0 i < K, 0 i + j < K) p(i, j; i, j 1) = jH (0 i K, 0 < i + j C) p(i, j; i, j + 1) = H (0 i K, 0 i + j < C). Although the dwell time in the stop phase may be not sufciently long, the CTMC model for the CP scheme will arrive at the stationary state if the vehicle experiences a number of stop phases since it is ergodic. Also, the steady state probability stop (i, j) of the CP scheme can be obtained by an iterative algorithm [12]. Hence, we will use steady state probability to obtain NCBP, HCDP, and channel utilization. Then, in the stop phase, since a new call is blocked when the number of occupied channels is equal to or greater than K, NCBP is given by
K stop PN B = i=0 j=Ki Ci

We dene the channel utilization as the ratio of the number of used channels to the total capacity. Then, the channel utilization is given by
K Ci

stop =
i=0 j=0

i+j stop (i, j). C

(3)

B. Moving Phase Figure 4 shows the two-dimensional CTMC model for the moving phase, which is the same as that of the nonprioritization (NP) scheme. As mentioned before, since there are no incoming handoff calls when the vehicle is moving, this CTMC model has only departure process in terms of handoff calls. Namely, the CTMC model for the moving phase is not ergodic and thus the steady state probability cannot be used. Hence, to obtain NCBP and channel utilization, we will regard the two-dimensional CTMC as n+1 one-dimensional CTMCs after inducing the probability that n handoff calls exist until the next stop phase. As shown in Figure 4, since new calls are rejected only when all channels in the WLAN are used up, the transition rates for new calls are given by p(i; i 1) = iN (0 < i C n) p(i; i + 1) = N (0 i < C n), where n is the number of handoff calls. Suppose that the vehicle goes into the moving phase when the state at the stop phase is (a, b). Also, denote that ta be the elapsed time for transition to the next stop phase, i.e., the average traveling time to the next stop. Then, the probability that a handoff call duration time, td , is larger than ta is Pr(td > ta ) = eH ta . Hence, the probability that n (0 n b) handoff calls exist after ta , Pn (ta ), can be

stop (i, j).

(1)

On the other hand, a handoff call is dropped only when the number of occupied channels is equal to the capacity C, and therefore HCDP can be obtained from
K stop PHD = i=0

stop (i, C i).

(2)

978-1-4244-5638-3/10/$26.00 2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.

computed as Pn (ta ) =
b

b (eH ta )n (1 eH ta )bn , n

(4)

where n=0 Pn (ta ) = 1. As mentioned before, the system can be characterized by a one-dimensional CTMC, which is a M/M/C n/C n (0 n b) queueing model. Then, NCBP in the moving phase when there are b handoff calls after the stop phase can be obtained as
b moving (b) PN B

=
n=0

Pn (ta ) B(C n).

(5)
Fig. 5. New call blocking probabilities versus new call load.

where B(s) is a well-known Erlang loss formula in the M/M/s/s queueing system. However, Eq. (5) is derived under the assumption that the state after the stop phase is (a, b). Hence, we need to consider for all possible states to compute the average new call blocking probability in the moving can be obtained moving phase. Then from Eq. (5), PN B as
CK moving = PN B b=0 C moving PN B (b) K

stop (i, b)
i=0 Cb

+
b=CK+1

moving PN B (b) i=0

stop (i, b) .

(6)

moving (i) be the steady state i of the M/M/s/s Let s queueing system and it is given by [13] moving (i) = s

s k=0

i /i! k /k!

0 i s,

where is the trafc load of the M/M/s/s system. Then, the channel utilization when there are b handoff calls after the stop phase, moving (b), can be derived as
b Cn

moving (b) =
n=0 i=0

Pn (ta )

i + n moving Cn (i). C

(7)

Similar to Eq. (6), by considering all possible states, moving can be obtained from
CK K

moving =
b=0 C

moving (b) moving (b)

stop (i, b)
i=0 Cb

+
b=CK+1

stop (i, b) .
i=0

(8)

V. N UMERICAL R ESULTS In this section, we evaluate the performance of the MACAC algorithm against the CP and NP schemes in terms of new call blocking probability (NCBP), handoff call dropping probability (HCDP), and channel utilization. For numerical results, let C = 30, K = 25, and ta = 600. The mean arrival rates of new and handoff calls, N and H , are both set to

1/20. The stop phase is independent of the moving phase and the stop phase is analyzed with the assumption of the steady state. Thus, the MA-CAC algorithm and the CP scheme have the same NCBP, HCDP, and channel utilization. In addition, the performance comparison between the CP and NP schemes is well-known. Therefore, our analysis focuses on the moving phase. Figures 5 and 6 demonstrate NCBP and channel utilization as a function of the new call load, respectively. In Figures 5 and 6, the handoff call load, H /H is set to 30 whereas the new call load N /N varies from 5 to 60. From Figure 5, it is observed that NCBPs of MA-CAC, NP, and CP schemes increase as the new call load increases. Also, NCBP of the CP scheme is the highest since the CP scheme exclusively reserves a portion of channel for handoff calls. On the other hand, the NP scheme has lower NCBP than MA-CAC since MA-CAC prepares guard channels for handoff calls during the stop phase and these handoff calls can occupy the channels while the vehicle is moving. However, it can be expected that NCBPs of MA-CAC and the NP scheme are converged as the average vehicular moving time, ta , becomes larger. Although it is not presented in Figure 5, NCBPs of both schemes have the same value when ta is equal to 6000. From Figure 6, the CP scheme has the lowest channel utilization since new calls cannot be accepted although there are some available channels. On the other hand, since the MA-CAC algorithm allows that all available channels can be assigned for new calls during the moving phase, the channel utilization of MACAC is higher than that of the NP scheme. Similarly to NCBP, channel utilizations of both MA-CAC and the NC scheme will be identical as ta decreases. Figures 7 and 8 show NCBP and the channel utilization as the handoff call load increases. Note that since the mean arrival rates of handoff calls are xed, the handoff call duration time varies to adjust the handoff call load. Also, the new call load N /N is set to 30. Figures 7 and 8 exhibit similar trends with Figures 5 and 6, respectively. However, from Figure 7, it can be found that NCBPs are slightly high even under low handoff call loads unlike Figure 5 since the new call load is xed at 30.

978-1-4244-5638-3/10/$26.00 2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.

VI. C ONCLUSIONS In this paper, we proposed the MA-CAC algorithm in vehicular WiFi networks, which introduces novel call admission control schemes depending on the mobility. That is, guard channels are assigned only at the stop phase to prioritize handoff calls while no guard channels are assigned during the moving phase to maximally utilize the WLAN capacity. Numerical results demonstrate that the MA-CAC algorithm can lower handoff call probability while maintaining high channel utilization. In our future work, we will consider more diverse features in vehicular WiFi networks, e.g., traveling time, user population, and so on. ACKNOWLEDGEMENT
Fig. 6. Channel Utilization versus new call load.

This work was supported in part by NRF of Korea Grant funded by the Korean Government (2009-0064397) and in part by the ITRC support program supervised by the NIPA (NIPA2010-C1090-1011-0004). Corresponding author of this paper is Sangheon Pack (shpack@korea.ac.kr). R EFERENCES
[1] S. Pack, H. Rutagemwa, X. Shen, J. Mark, and L. Cai, Performance Analysis of Mobile Hotspots with Heterogeneous Wireless Links, IEEE Transactions on Wireless Communications, vol. 6, no. 10, pp. 3717-3727, October 2007. [2] S. Pack, X. Shen, J. Mark, and L. Cai, Throughput Analysis of TCP Friendly Rate Control in Mobile Hotspots, IEEE Transactions on Wireless Communications, vol. 7, no. 1, pp. 193-203, January 2008. [3] Y. Kim, J. Ko, and S. Pack, A Measurement Study on Internet Access in Vehicular Wi-Fi Networks, submitted. [4] N. Dimitriou and R. Tazazolli, Quality of Service for Multimedia CDMA, IEEE Communications Magazine, vol. 38, no. 7, pp. 88-94, July 2000. [5] M. Ahmed, Call Admission Control in Wireless Networks: A Comprehensive Survey, IEEE Communications Surveys, vol. 7, no. 1, pp. 50-69, 4th quarter, 2005. [6] H. Liao, X. Wang, and H. Chen, Adaptive Call Admission Control for Multi-Class Services in Wireless Networks, in Proc. IEEE ICC 2008, May 2008. [7] S. Lee, K. Kim, K. Hong, D. Grifth, Y. Kim, and N. Golmie, A Probabilistic Call Admission Control Algorithm for WLAN in Heterogeneous Wireless Environment, IEEE Transactions on Wireless Communications, vol. 8, no. 4, pp. 1672-1676, April 2009. [8] I. Soto, C. Bernardos, M. Calderon, A. Banchs, and A. Azcorra, NEMO-Enabled Localized Mobility Support for Internet Access in Automotive Scenarios, IEEE Communications Magazine, vol. 47, no. 5, pp. 152-159, May 2009. [9] E. Navarro, A. Rad, and V. Wong, Connection Admission Control for Multiservice Integrated Cellular/WLAN System, IEEE Transactions on Vehicular Technology, vol. 57, no. 6, pp. 3789-3800, November 2008. [10] L. X. Cai, X. Shen, J. Mark, L. Cai, and Y. Xiao, Voice Capacity Analysis of WLAN with Unbalanced Trafc, IEEE Transactions on Vehicular Technology, vol. 55, no. 3, pp. 7520761, May 2006. [11] Y. Fang and Y. Zhang, Call Admission Control Schemes and Performance Analysis in Wireless Mobile Networks, IEEE Transactions on Vehicular Technology, vol. 51, no. 2, pp. 371-382, March 2002. [12] P. Lin, W. Lai, and C. Gan, Modeling Opportunity Driven Multiple Access in UMTS, IEEE Transactions on Wireless Communications, vol. 3, no. 5, pp. 1669-1677, October 2004. [13] L. Kleinrock, Queueing Systems: Theory. New York: Wiley, 1975, vol. I.

Fig. 7.

New call blocking probabilities versus handoff call load.

Fig. 8.

Channel utilization versus handoff call load.

In addition, the difference of NCBPs between MA-CAC and the NP scheme decreases when the handoff call load increases. This is because the total capacity is occupied by handoff calls under high handoff call load and thus the MA-CAC algorithm cannot allocate remaining resources to new calls.

978-1-4244-5638-3/10/$26.00 2010 IEEE

Você também pode gostar