Macrocell/Microcell Selection Schemes Based on a New Velocity Estimation in Multitier Cellular System

Transcription

1 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 51, NO. 5, SEPTEMBER Macrocell/Microcell Selection Schemes Based on a New Velocity Estimation in Multitier Cellular System Young-uk Chung, Student Member, IEEE, Dong-Jun Lee, Dong-Ho Cho, Senior Member, IEEE, and Byung-Cheol Shin Abstract The Intermational Mobile Telecommunications IMT-2000 system uses a microcell concept to provide multimedia services and to support increasing number of users. However, in a microcell system, the number of handoffs is greatly increased. To solve this problem, a multitier cellular structure is proposed, in which high-speed mobile terminals (MTs) are serviced in macrocell and low-speed MTs are serviced in microcell to minimize the number of handoffs. In this system, it is important to precisely estimate the speed of the MT the correct selection of the macrocell/microcell. We propose two macrocell/microcell selection schemes based on a new velocity estimation method in a multitier cellular system which uses the sojourn time in a microcell overlapping region. The proposed schemes have various advantages such as good permance when MT direction is varying, efficient user allocation to cells, quick velocity estimation capability, easy implementation, and low power consumption. We analyze and simulate the conventional and our proposed schemes in the Manhattan cell model, showing the proposed schemes have better permance than the conventional schemes. Index Terms Cell selection, multitier cell, sojourn time, velocity estimation. I. INTRODUCTION MOBILE communication systems in the future will provide multimedia services needing more channel resources per user. The number of users and traffic will greatly increase, resulting in a shortage of capacity in the current system using macrocells with a few kilometers diameter. To solve this problem, more capacity per frequency resource is required due to a limitation of usable frequency resources. The Intermational Mobile Telecommunications IMT-2000 system uses a microcell with diameter of a few hundred meters. Using such a microcell, we can increase the system capacity. However, in a microcell system, the number of handoffs increases greatly. The multitier cellular structure, which consists of microcell clusters superimposed on a macrocell, can solve this problem [1]. In this multitier structure, high-speed mobile terminals (MTs) are serviced in the macrocell while low-speed MTs are serviced in the microcells to minimize the number of Manuscript received November 27, 2000; revised January 14, 2002 and March 27, Y. Chung and D.-H. Cho are with the Department of Electrical Engineering and Computer Science, Korea Advanced Institute of Science and Technology (KAIST), Daejeon , Korea ( thatfox@mail.kaist.ac.kr; dhcho@ee.kaist.ac.kr). D.-J. Lee is with Samsung Electronics Co., Ltd., Suwon, Korea. B.-C. Shin is with the School of Electrical and Electronics Engineering, Chungbuk National University, Cheongju, Korea ( bcshin@cbucc.chungbuk.ac.kr). Digital Object Identifier /TVT handoffs. It is important to precisely estimate the speed of MT the purpose of proper selection of the macrocell/microcell. There are several previous works on channel assignment (or macrocell/microcell selection) in the hierarchical cellular systems, such as works based on user velocity and service properties. Most of the previous works are about channel assignment based on user velocity [2] [5]. These channel assignment schemes based on user velocity are classified into two issues: 1) how to estimate the velocity of users and 2) how to allocate channels to each user efficiently when it is assumed that the system knows the user velocity exactly. This paper is focused on the way to estimate velocity of users. In this paper, we estimate velocity of users and allocate channels based on the estimated velocity. In addition, research about how to estimate the velocity of users are classified into two main approaches such as the fading distribution property-based method and the cell sojourn time-based method [6] [19]. The proposed velocity estimation scheme belongs to the cell sojourn time-based method. There are three conventional schemes which belong to the cell selection method using sojourn time. The fading distribution property-based method gives a relatively accurate estimation in noiseless circumstances, but in noisy environment, results are unreliable [6] [16]. Thus, this fading distribution property-based schemes can be used in suburban area which is modeled as circular macrocell. But this scheme cannot be used in urban area which is modeled as Manhattan structured microcell, because the radio environment of the urban area is very noisy. On the other side, the cell sojourn time-based method works well in urban area, because the cell shape and structure is relatively regular. This method works without any added hardware, can be easily implemented, and gives a relatively accurate estimate. These two approaches can complement each other: Sojourn time-based schemes are used in the urban area and fading distribution property-based schemes are used in the suburban area. Moreover, there is one more difference between two methods: The fading distribution property-based method estimates the real velocity of an MT, while the cell sojourn time-based method determines whether an MT has a high or low speed. Using these velocity estimation methods, we can perm macrocell/microcell selection in a multitier cellular structured system. We use the cell sojourn time-based method. The cell selection method using sojourn time can be classified into three schemes: the residual dwell time (RDT) scheme [17], the power level offset (PLO) scheme [18], and the exponential averaging (EA) scheme [19]. All schemes select cells by comparing the /02$ IEEE

2 894 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 51, NO. 5, SEPTEMBER 2002 sojourn time with a predefined threshold time. The basic difference among these three schemes is the sojourn time measurement region. RDT scheme uses the residual microcell sojourn time of the MT after call origination. The call is handed off to a microcell or a macrocell based on the amount of residual time in the originating cell prior to a handoff. PLO scheme applies a specific negative offset to the received power level a specified time period when the MT is entering into a microcell and selects a macrocell/microcell by comparing the call origination time with the specified time period. In EA scheme, the MT tracks its microcell sojourn time even when it is idle. Cell selection is based on the estimated local mean of the sojourn times. We propose two new macrocell/microcell selection schemes using the cell sojourn time. The proposed schemes select cells based on the sojourn time in a microcell overlapping region. Using this region, we can reduce the sojourn time measuring region. This makes the proposed schemes less sensitive to direction change of an MT. Several attendant advantages are also present. Comparison of analysis and simulation results shows that the proposed schemes have better permance. In this paper, we compared permance of our proposed scheme with only three schemes, such as RDT, PLO, and EA scheme, because of fair comparison. This paper is organized as follows: In Section II, we describe existing schemes. Section III explains our proposed schemes. Analysis and simulation of the permance of each scheme and comparison of the results are presented in Section IV. In Section V, a numerical comparison is presented and in Section VI, we summarize our results and conclude this paper. II. EXISTING SCHEMES A. RDT Scheme In the RDT scheme, all calls are first assigned to a microcell [17], as shown in Fig. 1. After a call is established, the MT starts the measuring of the residual dwell time from the time when the call is originated to the time when the call is handed off. If the residual dwell time is longer than the predefined threshold time, the MT is determined to have low speed and handoffed to a microcell. Otherwise, the MT is determined to have high speed and handoffed to a macrocell. However, the call holding time of the MT in the call origination cell does not accurately reflect the speed of the MT, because the call holding time depends on the location where the call was actually originated in the microcell. Although this scheme can be easily implemented, the probability of erroneous assignment to a macrocell/microcell is high. B. PLO Scheme The PLO scheme selects a macrocell/microcell in idle mode [18]. When an MT enters into a microcell, a specific negative offset to the received power level from microcell base station (BS) is applied a specified time period, as shown in Fig. 2. If an MT originates a call during the specified time period, the call is established in the macrocell because of the negative offset. If a call is originated after the specified time period, the negative offset is removed and the call is established in the microcell. Fig. 1. Fig. 2. Model of the RDT scheme. Model of the PLO scheme. This scheme has been used the globel system mobile communications (GSM) system. Because the occurrence of call arrival is independent of the microcell boundary, the distribution of the sojourn time in a cell prior to call origination is the same as the residual dwell time in a cell after call origination. Theree, the probability of erroneous assignment in this scheme is the same as that in the RDT scheme. This scheme offers easy implementation, but suffers a high probability of erroneous assignment to a macrocell/microcell. C. EA Scheme In the EA scheme, the past microcell sojourn times of the MT are averaged [19]. The mean sojourn time is used macrocell/microcell selection. When an MT enters into a microcell, the sojourn time is measured until it is moved to another microcell as shown in Fig. 3. The MT repeats this procedure and averages the sojourn times using an exponential averaging method until a call is originated. Then, the MT compares the mean cell sojourn time with the predefined threshold. If the mean is smaller than the threshold, the call is established in a macrocell. If the mean is larger than the threshold, the call is established in a microcell. This scheme can select cells relatively accurately but suffers from complex implementation. The big disadvantage of this scheme is that it requires the microcell/macrocell overlay to be continuous throughout the system a sufficient distance such that the averaging mechanisms can converge. For an isolated area of a few microcells, this scheme fails miserably. Also, in case an MT is powered on and immediately initiates a call, this property is a fatal disadvantage because measurements of past

3 CHUNG et al.: MACROCELL/MICROCELL SELECTION SCHEMES 895 Fig. 4. Overlapped region model. Fig. 3. Model of the EA scheme. microcell cell sojourn times will not exist. Thus, an estimation of velocity can not be made and this scheme allocates calls randomly to a macrocell/microcell. This scheme also suffers from a power problem. Sojourn time is measured while the MT is in idle mode. However, the MT should maintain the status of active mode a while whenever it leaves a cell, to calculate the average measured sojourn time. The active mode is the state of MT after call origination, and the idle mode is the state of MT bee call origination. III. PROPOSED SCHEMES We propose two new macrocell/microcell selection schemes in a multitier cellular system. They are overlapping region sojourn time (ORST) scheme and averaging ORST (AORST) Scheme. We assume that the system being considered is a code division multiple access (CDMA)-based IMT-2000 system. In this system, each microcell and macrocell operate on separate frequencies. A. ORST Scheme The ORST scheme selects cells based on the sojourn time in a microcell overlapping region, as shown in Fig. 4. The algorithm of detecting the sojourn time measuring region is the same as the algorithm of detecting the soft handoff region in conventional soft handoff scheme, using the received pilot signal strength [20], [21]. As an MT moves toward the cell boundary, the received pilot signal strength from source cell BS deteriorates while the pilot signal strength from a neighboring cell BS becomes stronger. When the received pilot signal strength of neighboring cell is stronger than, we assume that the call enters into the overlapping region. After that time, if the received pilot signal strength from the current cell is less than and this status is maintained during,we assume that the MT leaves the overlapping region. This procedure is shown in Fig. 5. The values of,, and are threshold values defined in the CDMA standard. In practice, the overlap region of each cell can be different in real environment. So, each BS should have the inmation of its overlap region. This inmation may be included in initial setting of the BS. In addition, an MT can traverse regions with multiple microcell overlap regions, even though this case rarely occurs in the Manhattan microcellular environments [22]. In this Fig. 5. Algorithm of measuring ORST. case, the MT receives several pilot signals from nearby microcell BSs. In this algorithm, the MT determines source microcell and target microcell using the strength of pilot signal received from the BS. The microcell sending the strongest pilot signal except source microcell is determined to be the target microcell. Calls are first assigned to a microcell. After a call is established and the MT enters into the overlapping region, the measurement of the sojourn time begins. When a call is originated in the overlapping region, the MT measures sojourn time until it arrives at the boundary of the overlapping region. In the IMT-2000 system, the MT receives several pilot signals from nearby microcell BSs. When a call is originated, the microcell sending the strongest pilot signal is determined to be the source microcell. The microcell sending the second strongest pilot signal is determined to be the target microcell. The target microcell can be changed if the strength of pilot signals received from the microcell BS varies. The source microcell is not changeable. Cell selection is determined when the MT arrives at the boundary of the source microcell. After cell selection, the MT is handoffed to the target microcell or the locating macrocell. If the measured ORST is longer than the threshold time, the MT is estimated to have low speed and handoffed to a microcell. Otherwise, the MT is estimated to have high speed and handoffed to a macrocell. This proposed scheme has many advantages. First, we can get good permance not only in case that MT moves without changing direction, but also in case that direction of the MT is varying. When a fast-moving MT changes its direction, the sojourn time of the MT will be long and the MT is regarded as having a low speed. The smaller the sojourn time measuring

4 896 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 51, NO. 5, SEPTEMBER 2002 region, the lower the probability of changing direction in that region, especially in the Manhattan cell structure. This scheme also efficiently allocates user traffic to macrocell/microcells. The capacity of a macrocell is much smaller than the total capacity of its embedded microcells in the multitier structure. There can be fast moving MTs not requiring handoff because of short call duration. By cing MTs to select cells when they require handoff, only fast-moving MTs which will require handoff are allocated to the macrocell. This will minimize the macrocell blocking probability. In case that a user turns on an MT and immediately initiates a call, the MT velocity can not be estimated in the EA scheme. Thus, the selection of a macro/microcell is not possible. (See Section II-C.) However, the ORST scheme estimates velocity after a call is originated and selects a cell when handoff is required. Thus, the ORST scheme can select a cell normally in this case. Similarly, the ORST scheme works well in the multitier system which consists of a macrocell and a few isolated microcells. For an isolated area of a few microcells, the EA scheme fails miserably. The proposed ORST scheme can also be implemented easily, and requires only low battery power by reduction of the MT load. However, when the speed of the MT changes quickly, this scheme suffers a disadvantage because the sojourn time measuring region is small. B. AORST Scheme The AORST scheme also selects a cell based on the sojourn time of overlapping region, similar to the ORST scheme. However, in this scheme, the MT averages the ORST values macrocell/microcell selection, similar to conventional EA scheme. When an MT enters into a microcell, the overlapping region sojourn time is measured until the MT leaves the region. The MT repeats this procedure and averages the sojourn times using an exponential averaging method until a new call is originated. When a new call is originated, the MT compares the mean cell sojourn time with the predefined threshold. If the mean is smaller than the threshold, the call is originated in a macrocell. Otherwise, the call is originated in a microcell. This scheme has advantages similar to the ORST scheme (see Section III-A) and has similar advantages/disadvantages to EA scheme (see Section II-C). However, in comparison with EA scheme, the AORST scheme requires less time to get the same numbers of sample averaging, because this scheme measures ORST. For example, an MT in EA scheme must pass five cells to get five samples, but AORST scheme requires only four cells to get five samples. This advantage helps to overcome the weakness of the averaging method in requiring a relatively long time to get sufficiently good permance. IV. ANALYSIS AND SIMULATION We analyze and simulate the permance of each scheme comparison. To verify the efficiency of the proposed schemes, we perm analysis and simulation the Manhattan cell model [22]. The cell structure in a Manhattan cell model is Fig. 6. Manhattan cell model. shown in Fig. 6. A cell consists of several blocks and streets. We define the mobility model as follows. Cells are square with side length. Let be the minimum street length from a cell boundary to a corner and be the street length between two corners. Also, let be the total street distance in a microcell and be the length of an overlapping region. has the value of. MTs are unimly distributed on the center of streets. MTs move straight along streets and can turn at corners. At corners, MTs go straight with probability of, turn right with probability of, and turn left with probability of. We define that is a substitution notation of and,or. Return is not allowed. Speed of an MT does not change. In the analysis, we assume that an MT can change its direction maximum two times in a microcell. A. RDT Scheme and PLO Scheme The erroneous assignment probability of the PLO scheme is the same as the RDT scheme (see Section II-B). Theree, we analyze and simulate only the RDT scheme. Bee verifying the permance of RDT scheme, we must find the distribution of the residual dwell time. Let a call (MT) originate at some point in a microcell, where is the moving distance bee the MT leaves the microcell.

5 CHUNG et al.: MACROCELL/MICROCELL SELECTION SCHEMES 897 First, we perm the analysis in case that the MT moves without change of direction. Because the distribution of is unim from zero to, the probability density function (pdf) and the cumulative distribution (cdf) of can be found as (1) We also perm analysis in case that the MT moves with change of direction. The pdf and cdf of the distance can be described in (3) and (4) shown at the bottom of the page. The pdf and cdf are calculated by following the trace of an MT each case. Assume is the speed of an MT and is the residual (2) (3) where (4)

6 898 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 51, NO. 5, SEPTEMBER 2002 dwell time. Then, is given by. The pdf and cdf of are obtained by (5) Let and be the probabilities of erroneous assignment to a macrocell and a microcell, respectively. Given the threshold time, and can be obtained as MTs which have velocity below (9) MTs which have velocity above (10) Also, the mean of (6) can be calculated as Now, we can calculate the probability of erroneous assignment in case that a low-speed MT is allocated to a macrocell and a high-speed MT is allocated to a microcell. Let be the velocity threshold which decides whether the speed of an MT is fast or slow. Then, we can find the threshold time as follows: (7) B. EA Scheme In this model, the cell sojourn time is always definite in case that an MT does not change direction. Theree, the probability of erroneous assignment to cells is zero in this case. So, we analyze and simulate only in case that an MT changes direction. To verify the permance of this scheme, it is necessary to know the sojourn time distribution in a microcell. Let be the trajectory length of an MT in a microcell. Then, the pdf and cdf of are given by (11) (13) shown at the bottom of the page. The pdf and cdf are obtained by chasing the trace of an MT each case. The cell sojourn time is. Thus, the pdf and cdf of can be obtained as (8) (14) (11) (12) (13) where

7 CHUNG et al.: MACROCELL/MICROCELL SELECTION SCHEMES 899 (15) The mean and variance of are given by (16) Fig. 7. Overlapped region in Manhattan cell model. (17) Now we calculate the probability of erroneous assignment. Let be the threshold velocity which decides whether the speed of an MT is fast or slow. Then, we can find the threshold time as follows: (18) Let be a smoothing constant. If we calculate the average of samples using EA, is obtained as [23] C. ORST Scheme and AORST Scheme In the Manhattan cell model, the analyses of proposed schemes are the same whether the MT changes direction or not. Fig. 7 shows a simple rectangular overlapping region. Then, pdf, cdf, and the mean of (which is the moving distance in an overlapping region) are given by (25) The sojourn time in cell, is given by (19) (26) (20) where is a random noise sample with a mean and variance of zero and, respectively. Also, is the constant local mean sojourn time. The mean and variance of the estimated cell sojourn time are calculated as (27) From (16) and (17), and. From the central limit theorem, has a normal distribution [24]. Then, the pdf and cdf of are given by where. In (25) (27), the term indicates the probability of call origination in an overlapping region. Also, the term is the probability that a call originated in the overlapping region moves into a nonoverlapping region. Let the distance and velocity of the MT be independent random variables. Then, the mean of ORST is calculated as (21) (22) Let and be the probabilities of erroneous assignment to a macrocell and a microcell, respectively. Given the threshold time, and can be obtained as MTs which have velocity below (23) MTs which have velocity above (24) The variance of is obtained as Then, the pdf and cdf of are given by (28) (29)

8 900 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 51, NO. 5, SEPTEMBER 2002 (30) Let be the probability of erroneous assignment of a call to a macrocell. This means the probability of when the speed of the MT is slower than. Also, let be the probability of erroneous assignment of a call to a microcell. This means the probability of when the speed of an MT is faster than. Then,, and the threshold time are be given by MTs which have velocity below MTs which have velocity above (31) Fig. 8. model. Probability of erroneous assignment to macrocell in Manhattan cell where is the threshold velocity. For the analysis of the AORST scheme, we need more calculation. Let be a smoothing constant. Also, let be the sojourn time in cell and be a random noise sample with a mean and variance of zero and, respectively. If we assume that the velocity is constant,. Then, using the same calculation as shown in Section IV-B, we can get the pdf and cdf of (which is the estimated mean ORST) as Fig. 9. model. Probability of erroneous assignment to microcell in Manhattan cell (32) (33) where and. Then, and are be given by MTs which have velocity below MTs which have velocity above (34) D. Analysis and Simulation Results MTs can be vehicles or pedestrians. In the cell selection method based on sojourn time, all pedestrians are allocated to a microcell without any erroneous assignment because the speed of a pedestrian is almost zero and the sojourn time is almost infinite. Theree, we analyze and simulate only vehicles. In our analysis and simulation, we assume that the side length of a microcell is 560 m, a block is square with a side length of 160 m, the width of a street is 20 m, and the length of the overlapping region is 20 m. We also assume that m/s and MTs have a velocity from 1 m/s to 23 m/s. Let the speed of an MT not be changed. We find the probability of erroneous assignment each velocity. For the averaging scenario such as the EA and AORST schemes, we assume that an MT passes five microcells bee a new call is originated. We generate new calls which are unimly distributed only on the streets. MTs move straight along streets and can turn at corners. At corners, we assume that, and. Return is not allowed. We add more assumptions simulations. In the consideration of MT direction change, the direction varies after an exponentially distributed interval with a mean of 10 s. We assume there is no limitation the number of direction changes. In the simulation, we find the value which makes the probability of erroneous assignment the value of 0.5 when MT velocity is 12 m/s. This value is used as the threshold time to select cells. Analysis and simulation results are summarized in Figs. 8 and 9. Where an MT does not change direction, the EA scheme has no erroneous cell assignment. So, we exclude the results of EA scheme in this scenario. In this model, the ORST scheme has the same analytical and simulation results whether the MT does or does not change direction. This property is also shown in the AORST scheme. In these figures, A refers to the analytical result and S refers to the simulation result. Also, D indicates an MT that does not change direction while C indicates an MT that changes direction. is the probability that a slow MT

9 CHUNG et al.: MACROCELL/MICROCELL SELECTION SCHEMES 901 TABLE I SUMMARY OF THE PROBABILITY OF ERRONEOUS ASSIGNMENT Fig. 10. Distribution of mobile velocity. is allocated to a macrocell. This phenomenon gives rise to a shortage of macrocell capacity. is the probability that a fast MT is allocated to a microcell. This effect causes an increase in the number of handoffs. Based on these results, the proposed ORST and AORST schemes exhibit better permance than conventional RDT scheme and PLO scheme whether an MT changes direction or not. In comparison with EA scheme, two proposed schemes also have better permance. When an MT changes direction, proposed schemes exhibit better permance than conventional EA scheme as the velocity of the MT approaches the threshold velocity. The reason that proposed schemes have a probability of erroneous assignment is the existence of a new call originating in the overlapping region. In case that an MT does not change direction, EA scheme has no erroneous assignment. But the difference between proposed schemes and EA scheme is very tiny. Moreover, because changing direction is more close to the real environment, we can conclude that proposed schemes have the best permance in Manhattan cell model. V. NUMERICAL EXAMPLE Based on analysis and simulation results, we can calculate the blocking probability. Let a macrocell contain ten embedded microcells. Each macrocell and microcell has ten channels. Assume no channel reservation handoff calls. Let the call holding time be exponentially distributed with a mean of s. Let and be the call blocking probabilities in a macrocell and each microcell, respectively. Also, let and be the actual call arrival rates a macrocell and each microcell after cell selection, respectively. Let the load of a macrocell and each embedded microcell be the same. All pedestrian traffic is allocated to a microcell with 100% probability. Because most pedestrian terminal has almost zero speed, and the sojourn time of a zero-speed MT is infinity. We assume that the speed of a vehicle has a triangular distribution as shown in Fig. 10. Then, from analysis and simulation results in Figs. 8 and 9, the total and are calculated. The results are summarized in Table I. We calculate the blocking probability in case that the total call arrival rate and the ratio of the vehicular to the pedestrian traffic vary. For example, assume that the total call arrival rate to a macrocell and its embedded microcells is 1500 calls/h, and the ratio of the vehicular traffic to the pedestrian traffic is 2 : 1. Then, the call arrival rate of vehicular traffic is 1000 calls/h Fig. 11. Blocking probability in macrocell versus total call arrival rate. and the pedestrian traffic is 500 calls/h. The load on the macrocell is 140 calls/h and the load on each microcell is 136 calls/h, which consists of 86 calls/h vehicles and 50 calls/h pedestrians. From these assumptions, the actual call arrival rates to the macrocell and each microcell are given by (35) (36) and are obtained using the Erlang- mula. We calculate and only in case that an MT moves with direction change. We select this model because it most closely resembles a real environment. First, we calculate and while changing the total call arrival rate from 500 calls/h to 3000 calls/h. We assume the ratio of vehicular to pedestrian traffic is 2 : 1. Results show that the blocking probability increases as the total arrival rate increases, as shown in Figs. 11 and 12. In these figures, D indicates an MT that does not change direction, while C indicates an MT that changes direction. We calculate, in case the ratio of the vehicular to pedestrian traffic is changed. We assume the total call arrival rate is 1500 calls/h. Results show that the blocking probability in a macrocell decreases while the blocking probability in a microcell increases as the rate of pedestrian traffic increases as shown in Figs. 13 and 14. That is because, as the rate of pedestrian traffic increases, the traffic load on a macrocell decreases while the traffic load on a microcell increases. In Fig. 14, you can see the increasing trend of blocking probability in a microcell. But, the increasing range is very tiny compared to the decreasing range of blocking probability in macrocell. The reason

10 902 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 51, NO. 5, SEPTEMBER 2002 VI. CONCLUSION Fig. 12. Blocking probability in microcell versus total call arrival rate. Two cell selection schemes are proposed and described based on a new velocity estimation scheme in a multitier cellular system. The proposed schemes offer the advantages of detection of a direction-changing effect, efficient user allocation to cells, estimation of instant velocity, easy implementation, and low power consumption. Analysis and simulation of the probability of erroneous assignment to a macrocell/microcell in the Manhattan cell model were permed to verify that the proposed schemes exhibit good permance. Results the proposed schemes were compared with existing schemes. It is shown that the proposed schemes have better permance, especially in case of moving direction changed. Because moving direction is changed in the real environment, we can conclude that the proposed schemes have the best permance compared with conventional schemes. The blocking probability was also calculated when both the total call arrival rate and the traffic ratio vary. The proposed schemes have lower probability of erroneous assignment and better blocking permance than conventional schemes. REFERENCES Fig. 13. Fig. 14. Blocking probability in macrocell versus traffic ratio of pedestrian. Blocking probability in microcell versus traffic ratio of pedestrian. is because there are many microcells in the same area of one macrocell. Figs show that the proposed schemes have a lower blocking probability a macrocell and a higher blocking probability a microcell than the conventional schemes. However, the value of is almost zero and the difference is negligible all schemes. This means that more users are allocated to a microcell in the proposed schemes. Because the capacity of the macrocell is much smaller than total capacity of the embedded microcells, it is evident that the proposed schemes allocate user traffic to macrocell/microcells more efficiently than the existing schemes. [1] X. Lagrange, Multitier cell design, IEEE Commun. Mag., pp , Aug [2] C.-L. I, L. J. Greenstein, and R. D. Gitlin, A microcell/macrocell cellular architecture low- and high-mobility wireless users, IEEE Trans. Veh. Technol., vol. 11, pp , Aug [3] L. C. Wang, G. L. Stuber, and C. T. Lea, Architecture design, frequency planning, and permance analysis a microcell/macrocell overlaying system, IEEE Trans. Veh. Technol., vol. 46, pp , Nov [4] C. W. Sung and K. W. Shum, Channel assignment and layer selection in hierarchical cellular system with fuzzy control, IEEE Trans. Veh. Technol., vol. 50, pp , May [5] C. W. Sung and W. S. Wong, User speed estimation and dynamic channel allocation in hierarchical cellular systems, in Proc. IEEE Veh. Tech. Conf., 1994, pp [6] W. C. Jakes, Microwave Mobile Communications. New York: IEEE, [7] P. A. Bello, Some techniques the instantaneous real-time measurement of multipath and Doppler spread, IEEE Trans. Commun., vol. 13, pp , Sept [8] K. Kawabata, T. Nakamura, and E. Fukuda, Estimating velocity using diversity reception, in Proc. IEEE Veh. Tech. Conf., 1994, pp [9] A. Sampath and J. M. Holtzman, Estimation of maximum Doppler frequency handoff decisions, in Proc. IEEE Veh. Tech.Conf., 1993, pp [10] T. L. Doumi and J. G. Gardiner, Use of base station antenna diversity mobile speed estimation, Electron. Lett., vol. 30, no. 22, pp , [11] L. Wang, M. Silventoinen, and Z. Honkasalo, A new algorithm estimating mobile speed at the TDMA-based cellular system, in Proc. IEEE Veh. Tech. Conf., 1996, pp [12] K. Kawabata, T. Nakamura, and E. Fukuda, Estimating velocity using diversity reception, in Proc. IEEE Veh. Tech. Conf., vol. 1, 1994, pp [13] M. Hellebrandt, R. Mathar, and M. Scheibenbogen, Estimating position and velocity of mobiles in a cellular radio network, IEEE Trans. Veh. Technol., vol. 46, pp , Jan [14] J. M. Holtzman and A. Sampath, Adaptive averaging methodology handoffs in cellular systems, IEEE Trans. Veh. Technol., vol. 44, pp , Feb [15] R. Narasimhan and D. C. Cox, Speed estimation in wireless systems using wavelets, IEEE Trans. Commun., vol. 47, pp , Nov [16] L. Ehrman and R. Esposito, On the accuracy of the envelope method the measurement of Doppler spread, IEEE Trans. Commun., vol. COM-17, pp , Oct

11 CHUNG et al.: MACROCELL/MICROCELL SELECTION SCHEMES 903 [17] W. M. Jolley and R. E. Warfield, Modeling and analysis of layered cellular mobile networks, Teletraffic Datatraffic Period Change, vol. ITC-13, pp , [18] Unitel, Idle mode cell reselection microcell, in Proc. ETSI GSM2, Ronneby, Sweden, Sept [19] K. L. Yeung and S. Nanda, Channel management in microcell/macrocell cellular radio systems, IEEE Trans. Veh. Technol., vol. 45, pp , Nov [20] TIA/EIA-95-B, Mobile Station, Base station compatibility standard dual-mode spread spectrum systems,, Feb [21] 3GPP2 C.S0005-A,, Upper Layer (Layer 3) signaling standard cdma2000 spread spectrum systems,, June [22] M. Gudmundson, Cell planning in Manhattan environments, in Proc. IEEE Veh. Tech. Conf., 1992, pp [23] W. Gilchrist, Statistical Forcasting. Englewood Cliffs, NJ: Prentice- Hall, [24] A. Papoulis, Probability, Random Variables, and Stochastic Processes, 3rd ed. Englewood Cliffs, NJ: Prentice-Hall, Dong-Ho Cho (M 85 SM 00) received the B.S. degree in electrical engineering from Seoul National University, Seoul, Korea, in 1979 and the M.S. and Ph.D. degrees in electrical engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea, in 1981 and 1985, respectively. From 1987 to 1997, he was a Professor with the Department of Computer Engineering at Kyunghee University, Suwon, Korea. Since 1998, he has been with KAIST, where he is a Professor with the Department of Electrical Engineering and Computer Science. His research interests include wired/wireless communication network, protocol, and services. Young-uk Chung (S 98) received the B.S. and M.S. degrees in electrical engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea, in 1997 and 1999, respectively. Since 1999, he has been working toward the Ph.D. degree in electrical engineering and computer science at KAIST. His current research interests include resource management and location/mobility management in next-generation mobile communication systems. Dong-Jun Lee received the B.S., M.S., and Ph.D. degrees in electrical engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 1994, 1996, and 2000, respectively. Since 2000, he has been with Samsung Electronics Co., Ltd., Suwon, Korea. His research interests include CDMA systems, resource management, location management, and communication protocol in wireless networks. Byung-Cheol Shin received the B.S. degree in electrical engineering from the Seoul National University, Korea, in 1975, and the M.S. and Ph.D. degrees, both in electrical and electronics engineering, from the Korea Advanced Institute of Science and Technology (KAIST), Daejon, Korea, in 1977 and 1984, respectively. From 1977 to 1980, he was a Research Staff Member with the Electronics and Telecommunications Research Institute, Daejon, Korea. In 1987, he visited SRI International, Menlo Park, CA, as an International Fellow. In 1995, he also visited Rutgers University, New Brunswisk, NJ, as a Visiting Scholar. From 1984 to 1998, he was with KAIST. Since August 1998, he has been with Chungbuk National University, Cheongju, Korea, as a Faculty Member in the School of Electrical and Electronics Engineering. His research interests include real-time/multimedia communication and permance analysis of ATM, high-speed networks, and wireless networks.