用於智慧型運輸系統之CDMA多服務階層式細胞網路之效能分析

(1)

行政院國家科學委員會專題研究計畫成果報告

用於智慧型運輸系統之 CDMA 多服務階層式細胞網路之效能分析

計畫類別：個別型計畫

計畫編號： NSC94-2213-E-011-019-

執行期間： 94 年 08 月 01 日至 95 年 07 月 31 日執行單位：國立臺灣科技大學電機工程系

計畫主持人：鍾順平

計畫參與人員：蔡江鴻、林威廷、高逸竹、張益禎

報告類型：精簡報告

處理方式：本計畫可公開查詢

中華民國 95 年 10 月 2 日

(2)

行政院國家科學委員會補助專題研究計畫成果報告

支援非對稱性流量之多服務細胞網路之效能分析

Performance Evaluation of Hierarchical Multiservice CDMA Networks for Intelligent Transportation Systems

計畫編號：NSC 94－2231－E－011－019 執行期限：94 年 8 月 1 日至 95 年 7 月 31 日

主持人：鍾順平國立台灣科技大學電機系

計畫參與人員：蔡江鴻、林威廷、高逸竹、張益禎國立台灣科技大學電機系 一、中文摘要

我們提出一種用於智慧型運輸系統且支援語音與視訊流量之階層式細胞式 CDMA 網路。我們假設語音連結具有較短之服務時間，

而視訊連結具有較長之服務時間。視訊連結被指配給巨細胞以降低交遞之次數，而語音連結則指配給微細胞。我們利用保護通道來使交遞連結之優先權高於新連結。為提高資源使用效率，位於微細胞之語音連結可溢流至巨細胞。

為了避免過度之溢流，當情況允許時，位於巨細胞之語音連結將被取回至微細胞。我們利用多維馬可夫鏈來推導微細胞與巨細胞之數學模型。我們用 C 語言來撰寫模擬程式，以驗證分析結果的精確度。

關鍵字：階層式網路、智慧型運輸系統、服務時間、溢流、取回、保護通道

Abstract

We propose a hierarchical cellular CDMA network for ITS (Intelligent Transportation System) supporting both voice and video calls. It is assumed that voice calls have shorter average call holding time, whereas video calls have longer average call holding time. The video calls are assigned to the associated macrocell in order to reduce the number of handoffs and the voice calls are assigned to the associated microcell.

Guard channels are used to prioritize handoff calls over new calls. To increase utilization, voice calls in microcells can be allowed to overflow to macorcells. To avoid excessive overflow, overflowed voice calls are to be taken back to microcells if possible. Multi-dimensional Markov chains are used to describe microcells and macrocells, respectively. Last but not least, simulation programs written in C are run to verify the analytical results.

Key word: hierarchical networks, ITS, call

holding time, overflow, takeback, guard channel.

二、緣由與目的

In the Intelligent Transportation Systems (ITS), wireless communication is necessary to provide efficient management of transportation systems, such as electronic toll collection, vehicle location and navigation services [1]. It is expected that future ITS should support not only voice calls, but also multimedia calls.

As is well known, cellular concept has made mobile communication affordable to the general public. Thus, how to apply cellular concept to ITS has been studied in [1][5]. In [1], microcellular structures based on CDMA are proposed to support various services in ITS, where soft handoff scheme is adopted to provide seamless handoff. Soft handoff schemes have been studied in [1]-[3][7].

In previous works hierarchical cellular systems have been proposed to solve the unequal forced termination probability issue where mobile users have different moving speeds [6].

Although in the ITS environment the speed differences of mobile users are normally not significant, the diverse call holding times still can lead to unequal forced termination probability. In our work we propose a two-layer cellular network supporting both voice calls and video calls. It is assumed that voice calls have shorter average call holding time, whereas video calls have longer average call holding time. The video calls are assigned to the associated macrocell in order to reduce the number of handoffs, and thus the forced termination probability, whereas the voice calls are assigned to the associated microcell. Furthermore, to

(3)

enhance resource utilization in macrocells, voice calls may be allowed to overflow to the associated macrocell if they cannot find sufficient channels in the associated microcell [4]. However, excessive overflow may lead to QoS degradation of video calls in the upper layer.

Therefore, overflowed voice calls have to go back to the lower layer whenever there are enough channels available in the associated microcell, i.e., takeback [8]. It is also noted that generally forceful termination of an active call is less desirable than the blocking of a new call.

Thus, handoff calls are usually given priority over new calls.

To summarize, we consider hierarchical multiservice cellular networks for ITS based on CDMA supporting soft handoff. Both voice and video calls are supported in the considered system. Handoff calls are given priority over new calls. Three models are compared: no overflows, overflow without takeback, and overflow with takeback. An analytical method is derived to calculate the performance measures of interest.

三、結果與討論

3.1 System Model

The system studied covers a one- dimensional service area, where the mobile users can only move in one direction or the other.

There are two classes of users: class-1 (voice) and class-2 (video). It is assumed that voice calls have shorter average call holding time, whereas video calls have longer average call holding time.

To accommodate mobile users with different call holding times, a two-tier cellular network is proposed to implement the ITS., where each macrocell in the upper layer overlays N microcells in the lower layer. It is assumed that there areC⁽^M⁾(C^{( )}^m ) codes or channels in each macocell (microcell) It is noted that each call needs to be supported by one channel.

To enhance the seamless communication, soft handoff scheme is adopted, instead of hard handoff. Specifically, in each of microcells and macrocells, there are two regions that overlap with corresponding neighbor cells, which are referred to as soft handoff regions (SHR). The

non-overlapping region is referred to as normal region (NR). The ratio of the area of one SHR to the whole area of a cell is assumed to be f , and that of the area of NR to the whole area of a cell is 1 2 f− . One call in NR can connect to at most one base station (BS), normally the closest BS, whereas one call in SHR can connect to at most two BSs, the serving BS and the target BS.

Throughout this work, all of the parameters related to the microcells (macrocells) are represented with superscript m (M).

To provide almost equal forced termination probability and reduce the number of handoff attempts, voice calls are assigned to and served by the associated microcell and video calls by the associated macrocell. For further improving the Grade of Service (GoS) of voice calls, the voice calls that cannot obtain one free channel from the associated microcell can overflow to the upper layer and try to obtain the required channel from the associated macrocell. To avoid excessive overflow traffic and thus degradation of video calls’ GoS, the overflowed calls are regulated with the takeback process as described below.

With soft handoff, when a mobile station (MS) moves into SHR from NR, the MS will request the required resource from the target cell, while maintaining connected to the current cell until it moves out of the SHR into the NR of the target cell. Furthermore, we assume that when any voice call arrives at SHR^{( )}^m due to handoff or call initiation, it must obtain one free channel from each of the serving cell and the target cell;

otherwise it should release the channel obtained from the microcell and tries to overflow to macorcell. Similar policy applies to video calls.

In order to prioritize handoff calls over new calls, we also limit the accessible channels of the new calls by handoff guard channel mechanism.

3.2 Traffic Model

New call arrivals at a microcell (macrocell) are assumed to be Poison with average arrival rate λ₁for voice calls (λ₂ for video calls).

Handoff call arrivals for both voice and video calls are assumed to be Poisson. Overflow and takeback call arrivals for voice calls, if exist, are also assumed to be Poisson. The moving

(4)

direction of any call is generated at random and the direction is unchangeable during the call lifetime. The unencumbered call holding time of voice (video) calls is assumed to be exponentially distributed with mean 1/µ1

(

1/µ . 2

)

The cell dwell time is related to user mobility pattern and cell size. We assume that the microcell (macrocell) dwell time is also exponential with mean ^1/^µ^D^{( )}^m

(

^1/^µ^D⁽^M⁾

)

^{, and the}

cell dwell time is independent of the call holding time. Furthermore, we assume that the dwell times of the normal regions and soft handoff regions are exponentially distributed.

3.3 Call Admission Control

We consider three different models for the considered two-tier cellular network: Model A, Model B, and Model C. In Model A no overflows from the microcells to the macrocells are allowed, i.e., the two layers are independent of each other. In Model B voice calls in microcells can overflow to macrocells, if necessary, but no voice calls served by the macrocells can be taken back to microcells. In Model C voice calls in microcells can overflow to macrocells, if necessary, and voice calls served by macrocells should be taken back to the microcells whenever it is possible. It is noted that by choosing appropriate values for overflow guard channels and takeback guard channels, Model C can be reduced to Model B or Model A.

Specifically, if C_of =C_T⁽^M⁾, Model C reduces to Model A. Further, if C_of ≠C_T⁽^M⁾andC_tb =C_T^{( )}^m , Model C reduces to Model B.

3.4 Analytical Models

The state of a particular microcell is defined to be s^{( )}^m ={n1^{( )}_n^m,n1^{( )}_h^m}, where n₁^{( )}_n^m is the number of active voice calls at NR^{( )}^m , n₁^{( )}_h^m is the total number of incoming and outgoing active voice calls that are located at SHR^{( )}^m , where the incoming calls move toward the normal region of the serving cell, and the outgoing calls move away from the serving cell. A 2-dimensional Markov chain is used to describe the state transitions of a microcell. The associated equilibrium state equations are solved iteratively.

The performance measures of interest are found based on the steady-state probability distribution.

It is noted that in a macrocell, besides video new and handoff calls, there are new and handoff voice calls that overflow from microcells to macrocells due to insufficient resources in microcells. The state of a particular macrocell is defined to be s⁽^M⁾ =

( ) ( ) ( ) ( )

1 1 2 2

{n_n^M ,n_h^M ,n^M_n ,n^M_h }, where n₁⁽_n^M⁾ (n₂⁽^M_n⁾ ) is the number of active voice (video) call at

(M)

NR , n₁⁽_h^M⁾ ( n₂⁽^M_h ⁾ ) is the total number of incoming and outgoing active voice (video) calls atSHR⁽^M⁾. A 4-dimensional Markov chain is used to describe the state transitions of a macrocell. The associated equilibrium state equations are solved iteratively. The performance measures of interest are found.

3.5 Numerical Results

The numerical results, including both the simulation results and analytical results, are presented in this section. The simulation program is written in C language.

The effects of the new voice call arrival rate on various performance measures are studied with λ₂ =0.002. The new video call arrival rate is chosen so as to lead to P₂_B ≤10⁻² and

3

2_H 10

P ≤ ⁻ in Model A. According to these results, the analytical results are in reasonable agreement with the simulation results. The difference between analytical results and simulation results seems to appear in the heavy traffic conditions, where the analytical results overestimate simulation results.

Next, the effects of overflow and takeback on performance measures of interest are studied.

Recall that no overflows are allowed in Model A, only overflows are allowed but no takebacks in Model B, whereas both overflows and takebacks are allowed in Model C. Therefore, the performances of these three models are compared. First, it is observed that the blocking and forced termination probabilities of voice calls are lower in Model B and C than those in Model A. Obviously, this reduction in probabilities for voice calls results from the

(5)

overflow process. Specifically, in Models B and C the voice calls that cannot obtain the required channel in microcells, including both new call and handoff call arrivals, try to overflow to the associated macrocell and have a second chance to obtain the required channel, and this action results in better GoS for voice calls. It is also observed that the improvement of GoS for voice calls due to overflows is accompanied by the degradation of GoS for video calls. Comparing the performances of Models B and C, it is concluded that by allowing takebacks, Model C can achieve almost the same improvement in GoS for voice calls as Model B, while the GoS degradation for video calls in Model C is much smaller than that in Model B. The maximum allowable voice arrival rate under the constraints that new call blocking probability is 10⁻² and handoff failure probability is10⁻³ is considered, whereas the performance comparison for the three models withλ₁ =0.018 is performed to highlight the effects of overflow and takeback.

Obviously, the takeback process reduces the number of overflow voice calls in macrocells, resulting in smaller degradation of blocking and handoff failure probabilities of video calls when the voice traffic increases. The average numbers of occupied channels in a microcell and macrocell are studied, respectively. Models A and B result in the same channel utilization for a microcell. This is due to the fact that as long as a voice call cannot obtain the required channel from the associated microcell, it will leave that microcell, and specifically it will be rejected in Model A and it will overflow in Model B. Model C results in the highest channel utilization for a microcell among these three models, especially in heavy traffic load conditions This is obviously due to the fact that overflow voice calls may be taken back from macrocells to microcells.

Similarly, the channel utilization for a macrocell in Model C is lower than that in Model B, but it is still higher than that in Model A.

四、成果自評

We consider three models of a two-tier cellular CDMA network for ITS. The analytical method to compute the performance measures of interest for three models is derived with

multi-dimensional Markov chains. Both analytical and simulation results for performance measures of interest are presented. It is shown that for most of the scenarios considered the analytical results are in reasonable agreement with the simulation results. It is demonstrated that, compared with Model A, the overflow mechanism in Models B and C can achieve better GoS for voice calls at the expense of that for video calls. By comparing the performances of Models B and C, it is observed that the takeback mechanism in Model C can achieve almost the same GoS for voice calls as Model B, while Model C results in much better GoS for video calls than Model B.

五、參考文獻

[1] Y.-U. Chung and D.-H. Cho, “Performance Evaluation of Soft Handoff for Multimedia Services in Intelligent Transportation Systems Based on CDMA,” IEEE Trans. Intel. Transpor. Sys., vol. 4, no. 4, Dec. 2003, pp. 189-197.

[2] D. K. Kim and D. K.Sung, “Characterization of Soft Handoff in CDMA Systems,” IEEE Trans. Veh.

Tech., vol. 48 no. 4, July 1999, pp. 1195-1202.

[3] R. P. Narrainen and F. Takawira, “Performance Analysis of Soft Handoff in CDMA Cellular Networks,” IEEE Trans. Veh. Tech., vol. 50, no. 6, Nov. 2001, pp. 1507-1517.

[4] X. Wu, B. Mukherjee, and D.Ghosal, ”Hierarchical Architectures in the Third- Generation Cellular Network,” IEEE Wireless Comm., vol. 11, no. 3, June 2004, pp. 62-71.

[5] H. Harada, K. Sato, and M. Fujise, “A Radio-on- Fiber Based Millimeter-Wave Road-Vehicle Communication System by a Code Division Multiplexing Radio Transmission Scheme,” IEEE Trans. Intel. Transpor. Sys., vol. 2, no. 4, Dec. 2001, pp. 165-179.

[6] S-P Chung and J-C Lee, “Call-Holding-Time- Based Random Early Blocking in Hierarchical Cellular Multiservice Networks”, IEICE Trans. on Comm., Vol.E84-B No.4, Apr. 2001, pp.814-822.

[7] S-L Su, J-Y Chen, and J-H Huang, “Performance Analysis of Soft Handoff in CDMA Cellular Networks,” IEEE JSAC, vol. 14, no. 9, Dec. 1996, pp. 1762-1769.

[8] B. Li, C.-K.Wu, and Akira Fukuda, “Performance Analysis of Flexible Hierarchical Cellular Systems with a Bandwidth-Efficient Handoff Scheme,” IEEE Trans. Veh. Tech., vol. 50, no. 4, July 2001, pp.

971-980