A QoS-guaranteed fuzzy channel allocation controller for hierarchical cellular systems

1588 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 5, SEPTEMBER 2000

A QoS-Guaranteed Fuzzy Channel AllocationController for Hierarchical Cellular SystemsKuen-Rong Lo, Chung-Ju Chang, Senior Member, IEEE, Cooper Chang, and C. Bernard Shung

Abstract—This paper proposes a fuzzy channel allocationcontroller (FCAC) for hierarchical cellular systems. The FCACmainly contains a fuzzy channel allocation processor (FCAP)which is designed to be in a two-layer architecture that consistsof a fuzzy admission threshold estimator in the first layer and afuzzy channel allocator in the second layer. The FCAP chooses thehandoff failure probability, defined as quality-of-service (QoS)index, and the resource availability as input linguistic variablesfor the fuzzy admission threshold estimator, where the Sugeno’sposition gradient-type reasoning method is applied to adaptivelyadjust the admission threshold for the fuzzy channel allocator. Andthe FCAP takes the mobility of user, the channel utilization, andthe resource availability as input variables for the fuzzy channelallocator so that the channel allocation is finally determined,further based upon the admission threshold. Simulation resultsshow that FCAC can always guarantee the QoS requirement ofhandoff failure probability for all traffic loads. Also it improvesthe system utilization by 31.2% while it increases the handoff rateby 12.9% over the overflow channel allocation (OCA) scheme[7]; it enhances the system utilization by 6% and still reducesthe handoff rate by 6.7% as compared to the combined channelallocation (CCA) scheme [10], under a defined QoS constraint.

I. INTRODUCTION

DUE to the increasing demands for wireless communicationservices, it is essential to reconfigure the existing cellular

system into a hierarchical structure for enhancing system ca-pacity and improving coverage. The hierarchical cellular systemprovides overlaid microcells for high-teletraffic area and over-laying macrocells for low-teletraffic region [1]–[5].

In such a hierarchical cellular system, Rappaport and Hu pro-posed anoverflow channel allocation(OCA) scheme that al-lows a new or handoff call which has no channel available inthe overlaid microcell to overflow to use free channel in theoverlaying macrocell [6], [7]. The OCA scheme can reduce boththe blocking probability of new calls and the forced terminationprobability of handoff calls, and it is easy to implement becauseno elaborate coordination between microcells is needed [6]–[8].Beraldi et al. proposed areversiblehierarchical scheme [9],which allows the presence of handoff attempts from overlayingmacrocell to overlaid microcell if there is idle channel availablein the overlaid microcell. The reversible hierarchical scheme im-proves channel utilization in microcells and decreases blockingprobabilities of both new call and handoff call since microcells

Manuscript received November 8, 1998; revised January 13, 2000. This workwas supported in part by the National Science Council, Taiwan, under ContractsNSC 86-2213-E009-006 and NSC 87-2218-E009-047.

The authors are with the Department of Communication Engineering, Na-tional Chiao Tung University, Hsinchu, Taiwan.

Publisher Item Identifier S 0018-9545(00)07882-8.

are designed to be capable of supporting high capacity and bal-ancing the traffic load.

We also proposed acombined channel allocation(CCA)mechanism for hierarchical cellular systems [10]. It combinesoverflow, underflow, and reversible schemes, where new orhandoff calls having no idle channel to use in the overlaidmicrocell canoverflow to use free channels in the overlayingmacrocell, handoff calls from neighboring macrocell canunderflowto use free channels in the overlaid microcell, andhandoff attempts from macrocell-only region to a microcell inthe same macrocell can bereversedto use free channels in themicrocell. Simulation results showed that the CCA mechanismcan attain better channel utilization by an amount of 23.7% butrenders more handoff rate [11] by an amount of 19.8% than theOCA mechanism.

To guarantee QoS requirement for handoffs, these conven-tional techniques are to reserve a fixed number of guard chan-nels or to provide a queue for handoffs [1], [12]. However, thesemethods are unable to cope with burstiness of traffic. In otherwords, though these protection schemes can even guarantee theQoS requirement of the handoff failure probability, the channelutilization would suffer from fundamental limitations by unpre-dictable statistical fluctuations within new and handoff calls.And they are difficult, if not impossible, to drive an accuratemathematical model to obtain the solution.

On the other hand, fuzzy logic control has growing success invarious fields of applications, such as decision support, knowl-edge base systems, and pattern recognition. It is due to the in-herent capability of fuzzy logic control to formalize control al-gorithms that can tolerate imprecision and uncertainty, emu-lating the cognitive processes that human beings use every day[13]–[15].

In addition, when applied to problems, fuzzy systems haveoften shown a faster and smoother response than conventionalsystems. It is thanks to the fact that fuzzy control rules areusually simpler and do not require great computational com-plexity. The latter aspect, along with the spread of very largescale integration (VLSI) hardware structure dedicated to fuzzycomputation, makes fuzzy systems cost effective [16]. In thefield of telecommunications, fuzzy systems are also begin-ning to be used in areas such as traffic control in ATM net-works and channel allocation in mobile communication sys-tems [17]–[20].

In this paper, we propose a QoS-guaranteedfuzzy channelallocation controller(FCAC) for hierarchical cellular systems.The FCAC contains a fuzzy channel allocation processor(FCAP), a resource estimator, a performance evaluator, and

0018–9545/00$10.00 © 2000 IEEE

Lo et al.: QoS-GUARANTEED FUZZY CHANNEL ALLOCATION CONTROLLER 1589

Fig. 1. The fuzzy channel allocation controller for hierarchical cellular systems.

base-station interface modules. It dynamically estimates avail-able resources in macrocell and microcells, evaluates systemperformance, and determines whether and how to allocateresources to a call, based upon the call’s QoS requirement, re-source availability, and mobility. The FCAP is a two-layer fuzzylogic controller that contains the fuzzy admission thresholdestimator in the first layer and the fuzzy channel allocator in thesecond layer. In the fuzzy admission threshold estimator, weapply the Sugeno’s position-gradient type reasoning method toadaptively adjust the admission threshold for the fuzzy channelallocator. In the fuzzy channel allocator, we design to achieveutilization balancing between macrocell and microcells in orderto obtain a higher channel utilization. The domain knowledge isbased upon CCA mechanism we proposed in [10]. Simulationresults show that FCAC can guarantee QoS requirement ofexisting calls; also, it achieves better system utilization by anamount of 31.2% but more handoff rate by an amount of 12.9%than OCA, and it attains more system utilization by 6% andstill less handoff rate by 6.7% than CCA.

The rest of the paper is organized as follows. In Section II,functions of FCAC are described. Section III presents the designof FCAP. Section IV shows simulation results and discussions.Finally concluding remarks are given in Section V.

II. FUZZY CHANNEL ALLOCATION CONTROLLER(FCAC)

Fig. 1 shows the functional block diagram of the fuzzychannel allocation controller (FCAC) for hierarchical cellularsystems, where the hierarchical cellular system contains alarge geographical region tessellated by cells, referred to asmacrocells, and each of which overlays several microcells.The overlaying macrocell is denoted by cell 0 and its overlaidmicrocells are denoted by cell . For cell , a number

of channels is allocated, . FCAC containsfunctional blocks such as base-station interface module (BIM),performance evaluator, resource estimator, and fuzzy channelallocation processors (FCAP). It is installed in either a basestation controller (BSC) or mobile switching center (MSC).Note that for simplicity, FCAC is drawn to do the channelallocation for one macrocell only. Functional blocks of FCACare described as follows.

A. Base-Station Interface Modules (BIM)

BIM is to interface with the base station of macrocellor microcell. It provides complete partitioning buffers forqueueing new and handoff calls which are originated in thecorresponding cell and temporarily have no free channel to use.In the BIM for cell 0 ( ), there is a new-call buffer withcapacity for new calls originating in the macrocell-onlyregion, a handoff-call buffer with capacity for handoff callsfrom adjacent macrocells, an overflowed handoff-call bufferwith capacity for overflowed handoff calls from overlaidmicrocells, and an overflowed new-call buffer with capacity

for overflowed new calls from overlaid microcells. In theBIM for cell ( ), , there is a new-call bufferwith capacity for new call originations, an underflowedhandoff-call buffer with capacity for underflowed handoffcalls from the overlaying macrocell, and a handoff-call bufferwith capacity for handoff calls from adjacent microcells.No buffer is provided for the reversible handoff calls. Renegingof new calls and dropping of handoff calls are consideredbecause of new calls’ impatience and handoff calls’ movingout the handoff area.

Whenever receives a call request, , it sendsthe necessary calling information to the resource estimator, theperformance evaluator, and the FCAP. The calling information


TABLE ICALCULATION OF AVAILABLE RESOURCE

can distinctly indicate from which cell and in what type the callis originated. The type of call is defined as: denotes anew call originating in macrocell-only region; denotes ahandoff call from adjacent macrocell to macrocell-only region;

denotes a handoff call from microcell to macrocell-onlyregion; denotes new call originating in microcell;denotes handoff call from adjacent macrocell to an overlaid mi-crocell; denotes handoff call from microcell to microcell;and denotes reversible handoff. Note that the macro-cell-only region is the area inside macrocell0 but outside allmicrocells. The first three types of calls are to use channels inmacrocell, while other types of calls can use channels either inmacrocell or in microcell.

B. Resource Estimator

The resource estimator calculates the available resources inmacrocell0 and microcell when it receives calling informationof the type- call from at the time instant, denoted by

and , respectively.Since it knows system parameters of, , , ,

, and , , , , , it can obtainand by formulas shown in Table I.

In Table I, ( ) is the number of occupied channels in( ) at time ; ( , , ) is the number

of waiting calls in the new-call buffer (handoff-call buffer,overflowed handoff-call buffer, overflowed new-call buffer) of

, at time ; and ( , ) is the number ofwaiting calls in the new-call buffer (underflowed handoff-callbuffer, handoff-call buffer) of , , at time .

C. Performance Evaluator

The performance evaluator is to calculate the channel utiliza-tion and the handoff failure probability. The channel utilizationof macrocell0(microcell ) at time , denoted by ( ),is defined as

(1)

where ( ) is the average number of busy channels inmacrocell0 (microcell ) at time and ( ) is the channel ca-pacity for macrocell0 (microcell ). In order to show the channelutilization balancing between macrocell and microcells, we fur-

ther define a spatial averaging channel utilization of microcellsat time , denoted by , as

(2)

Also we define the system utilization of the whole system attime , denoted by , as

(3)

The handoff failure probability in macrocell (microcells) attime , denoted by ( ), is defined as

(4)

where( ) number of blocked waiting handoff

calls in macrocell0 (microcell ) attime ;

( ) number of dropped handoff calls inmacrocell0 (microcell ) at time ; and

( ) number of handoff calls in macrocell 0(microcell ) at time .

The handoff failure probability of the whole system at time,denoted by , is given by

(5)

D. Fuzzy Channel Allocation Processor (FCAP)

The FCAP performs the channel allocation using fuzzy logiccontrol to attain high channel utilization and keep the QoS re-quirement guaranteed. Here, a threshold is designed for channelallocation, and it can be adaptively adjusted to cope with theinput traffic fluctuation. The detailed design of an FCAP is de-scribed in the next section.

III. D ESIGN OFFCAP

As Fig. 2 shows, an FCAP mainly consists of a fuzzy admis-sion threshold estimator and a fuzzy channel allocator.

A. Fuzzy Admission Threshold Estimator

The fuzzy admission threshold estimator is to adaptively de-termine the decision thresholds for the fuzzy channel allocator


Fig. 2. The block diagram of fuzzy channel allocation processor.

so that the QoS requirement of handoff calls can be guaranteedand a high channel utilization can be achieved. The system QoSrequirement is here defined as the maximum handoff failureprobability which is denoted by . The admission thresholdsfor macrocell0 and microcell during time period , de-noted by and , are sent to the fuzzy channelallocator if a new call occurs in macrocell0 or microcell attime . Note that “ ” denotes the time instant the next newcall arrives, and “1” denotes one unit of a new-call interarrivaltime.

We choose ( ) and ( ) as input lin-guistic variables for the fuzzy admission threshold estimator todetermine ( ) for macrocell0 (microcell ).Since the fuzzy rule set will be commonly used by macrocell0 and microcell , hereafter in this subsection we simply use

, and , instead of or ,or , and or .

Term sets for input variables are designedas Low, High , and

Enough, Not Enough . And afunction defined below is chosen to betheir membership function. It is a trapezoidal function given by

for

for

for

otherwise

(6)

where ( ) is the left (right) edge of the trapezoidal function;( ) is the left (right) width of the trapezoidal function.Denote and to be the membership

functions for and in , respectively, and defineand as

(7)

(8)

would be set to be , and could be the safetymargin provided to tolerate the dynamic behavior of handofffailure probability and guarantee the QoS requirement, and theedge is .

Similarly, let and be the membershipfunctions for and in , respectively, and define

and as

(9)

(10)

The maximum possible “enough” value of available resourcewould be the sum of buffer size and allocation channels,

would be a safety margin of available resource, would beset to be a fraction of available resource, andand areprovided to tolerate the change of traffic.

Based on chosen input variables and their terms, the fuzzy ad-mission threshold estimator has fuzzy IF–THEN rules.The fuzzy rules have the following form:

Rule IF is and is

THEN

where is the term of theth linguistic variable used in rule, and is the output function of rule for the time

period , .Here, we apply the Sugeno’s position-gradient type reasoning

method [21], [22] to effectively derive and then toobtain . The inference in the Sugeno’s method has abuilt-in defuzzification such that can be expressed as

(11)

where, , forgetting constant that can maintain the esti-

mator stability,admission threshold during last time period

, andadjustment parameter for .


Then is given by

(12)

where is the weighting factor for the output variableof rule , defined as .Since membership functions are set to be symmetrical,

.The adjustment parameter is obtained by the gra-

dient decent method, where an error function at timeis definedas

(13)

Note that is given in the sense that needs to be con-trolled around . Then is given by

(14)

where is an adaptation gain which must be properly chosen.Because the admission threshold could be regarded

as an entry barrier to regulate new calls coming to the system,and the change of during , denoted by

would be varied in accordance with sothat can be kept at around to fulfill QoS require-ment, we heuristically make an approximation thathas a first-order relationship with . Then canbe expressed as

(15)

where is an experience value andis a constant value. Forexample, meansthat in the range of is inverselyvaried with respect to in the range of . Since

is designed to change gradually at a rate of ,the value of could be chosen to be . However, in order toattain a better adjustment of to fulfill QoS requirementin acute traffic fluctuation, we further set to be dynamicallychanged according to the number of handoff calls during

. In our example, it is frequent that only few handoffs occurduring , thus is set to be , where

is the number of handoff calls during .Equation (15) can be rewritten as

(16)

where denotes one unit of a new-call interarrival time. Andthen we have an expression for as

(17)

where is a pole which is set to a value very close to but lessthan one to avoid infinite memory and marginal stability of thisgradient evolution.

Finally, we rewrite (11) as

(18)

and we obtain by

(19)

B. Fuzzy Channel Allocator

We choose five input linguistic variables for fuzzy channelallocator: the channel utilization in macrocell 0 ( ), thechannel utilization in microcell ( ), the available resourcein macrocell0 ( ), the available resource in microcell( ), and the mobile speed (). and can show thetraffic load intensity and the load balancing among cells;and can implicate the remaining capacity; andcan showthe handoff rate. The term set for both and is de-fined as Low, High , theterm set for both and isEnough, Not Enough , and the term set for is

Slow, Fast . Let ( )and ( ) denote the membership functionsof terms and in ( ), respectively, and let

, , , and be

(20)

(21)

(22)

(23)

Here would be a fraction of channel utilization, and andwould be a change rate of channel utilization provided to

tolerate the dynamic behavior of and , respectively.The membership functions for terms ofand of and

have the same definition as in (9) and (10). The speed ofthe mobile user is hard to obtain. Usually the velocity of a mo-bile station can be estimated by the global positioning system(GPS), the Doppler effect, the elapsed time in a cell, the prop-agation time of the signal, and the number of level crossingsof the average signal level [23]. The membership functions forterms and in , denoted by and , are given by

(24)

(25)

where( ) would be a fraction of slow (fast) speed of mobile

user,( ) is provided to tolerate the change of slow (fast)

speed, andwould be the fastest speed.

There are two output variables, and , in the fuzzychannel allocator. The output variable represents whetherthe call is accepted or rejected, and indicates with which


channel in either macrocell or microcell the call is allocated.In order to provide a soft channel allocation decision, notonly “accept” and “reject” but also “weak accept” and “weakreject” are employed to describe the allocation decision.Thus, the term set for the output linguistic variable isdefined as Accept , Weak Accept ,Weak Reject , Reject , and the term set for isdefined as , Macrocell Microcell .

We further define a delta function, , for the member-ship functions of the output linguistic variables, whereis characterized as and ,

; , . The membership functionsfor terms , , , and in , denoted by , ,

, and , respectively, are given by

(26)

(27)

(28)

(29)

Without loss of generality, we set , , and let, . A call

can be allocated with channel if is greater than the admissionthreshold . And the membership functions for termsand in , denoted by and , respectively, aregiven by

(30)

(31)

and are set to be positive one and negative one, re-spectively. If is greater than zero, the call is allocated with amacrocell channel, otherwise, with a microcell channel.

There are different call types in hierarchical cellularsystems. For calls that can use only macrocell’s channel,only input linguistic variables of and are en-abled. For calls that could use channels in either macro-cell or microcell, input linguistic variables of ,

, , , and are enabled. The fuzzy rulebase with dimension is shownin Table II for the former, and that with dimension

is shown in Table III for the latter, where denotes thenumber of terms in .

The general idea of designing the fuzzy rules listed in TablesII and III is described as follows. If the available resource ineither macrocell or microcell is enough, a call would have achance of entering the system, andvice versa; if the availableresource is enough in macrocell but not enough in microcell,the macrocell channel would be preferred, andvice versa; wechoose a cell with low channel utilization, instead of the onewith high utilization, for balancing traffic load if the available

TABLE IIINFERENCERULES FORMACROCELL ONLY REGION

resource in both macrocell and microcell has same fuzzy termsin the premises of the fuzzy rule; and we also allocate calls tobe biased toward macrocell if the speed is fast for lesseningfrequent handoff, andvice versa.

Themax–mininference method is adopted. It first applies themin operator on membership values of the terms of all inputlinguistic variable for each rule. Assume that a call is originatedin microcell and the inference rules in Table III are applied.We denote the minimal result for ruleto be , ,and obtain , for example, by

(32)

Then the method applies themaxoperator to yield the overallmembership value. For the output variable, there are fourrules for term in Table III, which are rules 4, 8, 20, 28. Thenthe overall membership value for the term, denoted by ,is given by

(33)

Similarly, , and are yielded as

(34)

(35)

(36)

Afterwards, we use thecenter-of-areadefuzzification methodto derive the defuzzification value. The defuzzification value,denoted by , is given as shown in (37) at the bottom of thepage. Then the output variable is obtained by

forfor .

(38)

(37)


TABLE IIIINFERENCERULES FOROVERLAY REGION

Note that if the call is a new calland is originated in macrocell-only region, and

if the call is a new call and is orig-inated in microcell . On the other hand, in order to give a goodprotection for handoff calls, if the call is a handoff.

We similarly adopt themax–mininference method and applythe center-of-areadefuzzification method for output variable

, not further described here.

IV. SIMULATION RESULTS AND DISCUSSIONS

In the simulations, a hierarchical cellular system withmicrocells constructed along the Manhattan streets is assumed,

and the handoff behavior of users is characterized by a teletrafficflow matrix [4], defined as shown in the equation at the bottomof the page where , , represents the probability of ahandoff call originated in cell and directed to cell,

, and denotes the probability of this handoff call directedto the adjacent macrocell. for , and

would be zero.The number of mobile stations in each cell is assumed to be

550, and the new-call arrival process follows a Poisson processwith calling rate per mobile station (user). We assume thatlow- and high-mobility users are generated in a ratio of 7 : 3, andthe cell dwell time is exponentially distributed with mean 180 s(18 s) for the high-mobility users in macrocell (microcells) and


with 1440 s (144 s) for low-mobility users in macrocell (micro-cells). And the speed of mobile users is assumed to be uniformlydistributed in the range of 0–40 km (40–80 km) for low- (high-)mobility users. We also assume that the mean unencumberedsession duration is 100 s and the patience (dwell) time forqueued new (handoff) calls is in the range of 5–20 s. Thereare 150 channels fixedly allocated to macrocell and microcellswith a pattern of .If the OCA and CCA schemes are applied, the system re-serves a number of channels as guard channels forhandoff calls in cell , , which are denoted by

. We do some simulations and obtainthe appropriate forOCA scheme and forCCA schemes at . Since the reneging (dropping)process is considered, it is not necessary to provide a largebuffer size for new and handoff calls [10]; all buffer sizes inmacrocell and microcells are assumed to be 3. Note that in thefollowing performance comparisons, an OCA scheme providesno buffer and the CCA scheme supports the same bufferingscheme and capacity as FCAC does.

Based upon the QoS requirement and the knowledge of theCCA mechanism, parameters of membership functions for inputlinguistic variables in the fuzzy admission threshold estimatorare selected as follows: , , and

for and in (7) and (8);, , and for

and with macrocell in (9) and (10);, , and for

and with microcells in (9) and (10). In the fuzzychannel allocator, parameters of membership functions for inputlinguistic variables are selected as follows:

for , , , and in(20)–(23); , , and for

and with macrocell in (9) and (10);, , and for and

with microcells in (9) and (10); andkm, km, km, and km forand in (24) and (25). And constant parameters are set tobe and .

Three more performance measures such as the new-callfailure probability, the forced termination probability, and thehandoff rate are concerned, in addition to and .The new-call failure probability at time, denoted by ,is defined as

(39)

where ( ) is the number of blocked (reneging)new calls in cell and ( ) is the number ofnew calls originating in microcell (macrocell-only region), attime . Forced termination of a call occurs if a call is corrupteddue to a handoff failure during its conversation time. The

Fig. 3. P (t) andP (t) for FCAC, OCA, and CCA schemes.

forced termination probability at time, denoted by , isdefined as

(40)

where ( ) is the number of blocked (dropped)handoff calls in cell and is the number of admittednew calls originated in cell, at time . The handoff rate at time, denoted by , is defined as

(41)

Fig. 3 shows the new-call failure probability and thehandoff failure probability for schemes of FCAC, OCA,and CCA versus the calling rate per userat time .It can be seen that, asvaries, of FCAC remains con-stant at around , denoting the system QoS is guar-anteed, and is minimized so as to maximize the systemcapacity, comparing with OCA and CCA. It is because FCACuses fuzzy logic control and considers more system informationthan conventional schemes to allocate channels. Fuzzy logic isa soft logic as the truth value of an entity, not restricted to ei-ther false or true, but in a continuum of [0, 1]. And softness oftruth value is more appropriate to represent in determining if agiven requirement constraint is complied with or violated. Thisin effect removes the imposition of worse case assumption fromthe decision making of channel allocation. It is also because thefuzzy admission threshold estimator adopts the Sugeno’s posi-tion gradient-type reasoning method to effectively estimate the


Fig. 4. U(t), U (t), andU (t) for FCAC, OCA, and CCA schemes.

optimal admission threshold contained in each rule from its ob-served information; and the fuzzy channel allocator appropri-ately controls the admission of calls according to fuzzy admis-sion threshold value and allocates channels in either microcellor macrocell. While the conventional policies are inadaptive todetermine the number of guard channels to maintain, but not tooverprotect, the QoS requirement as the traffic load is fluctu-ating and the changing is unpredictable.

Fig. 4 shows the overall system utilization , the channelutilization in macrocell and in microcell versus thecalling rate per user for schemes of FCAC, OCA, and CCA attime . It reveals that of FCAC gains 31.2% and 6%improvement over the OCA and CCA methods, respectively;

of FCAC is increased by an amount of 8.4% over OCAbut is decreased by 2.1% under CCA, while of FCAC out-performs OCA and CCA by a significant amount of 44.6% and10%, respectively; and the pair of and for FCAC isthe closest one among those for OCA and CCA schemes. Thelatter two phenomena justify that FCAC can achieve the highestsystem utilization for more balancing utilization among cellsthan conventional schemes. And it is because fuzzy logic is apowerful tool that allows us to qualitatively represent controlrules naturally, on the basis of a simple linguistic description, toovercome some uncertainty and imprecision.

Fig. 5 shows the forced termination probability versusthe calling rate per userfor schemes of FCAC, OCA, and CCAat time . It is found that of FCAC has flat curvearound at . This is because we have obtained the unchanged

, shown in Fig. 3.Fig. 6 shows the handoff rate versus the calling rate per userfor schemes of FCAC, OCA, and CCA at time . It re-

veals that FCAC has more handoff rate by an amount of 12.9%than OCA. The reason is that the design of FCAC is based on theknowledge of CCA which combines overflow, reversible, andunderflow. However, the signaling overheads for these handoffsmight not cost as much as those for conventional handoffs be-tween macrocells since most of these handoffs occurred in thesame macrocell. And FCAC achieves a lesser handoff rate than

Fig. 5. P (t) for FCAC, OCA, and CCA schemes.

Fig. 6. R (t) for FCAC, OCA, and CCA schemes.

CCA. It is not only because of more information such as thespeed of mobile station considered in FCAC but also becauseof the fuzzy logic control that can provide decision support andexpert system with powerful reasoning capability.

V. CONCLUDING REMARKS

In this paper, we propose a QoS-guaranteed FCAC for hierar-chical cellular systems. The FCAC is designed to be a two-layercontroller which consists of a fuzzy admission threshold es-timator in the first layer and a fuzzy channel allocator in thesecond layer. The fuzzy admission threshold estimator appliesthe Sugeno’s position-gradient type reasoning method to adap-tively adjust the admission threshold value so that the QoS con-straint can be kept. The fuzzy channel allocator uses soft logic todetermine whether a call is accepted or not and which channelin macrocell or microcell will be allocated. Simulation resultsshow that the proposed FCAC improves the overall channel uti-lization 31.2% higher than the OCA scheme and 6% better than


the CCA scheme, while maintaining the QoS requirement; andit still reduces the handoff rate by an amount of 6.7% under theCCA mechanism but increases the handoff rate by an amountof 12.9% over the OCA mechanism. Since most of the hand-offs mentioned here are within the same macrocell, the signalingoverheads for these handoffs are not as much as those needed inhandoffs between macrocells. The FCAC would be a promisingand feasible approach.

ACKNOWLEDGMENT

The authors would like to thank the anonymous reviewersfor their valuable comments to improve the presentation of thepaper.

REFERENCES

[1] V. R. Kolavennu and S. S. Rappaport, “Traffic performance characteri-zation of a personal radiocommunication system,”IEE Proc. Commun.,vol. 133 F, pp. 550–561, Oct. 1986.

[2] C. L. I, L. J. Greenstein, and R. D. Gitlin,, “A microcell/macrocell cel-lular architecture for low- and high-mobility wireless users,”IEEE J.Select. Areas Commun., vol. 11, pp. 885–891, Aug. 1993.

[3] L. P. Chin and J. F. Chang, “Performance analysis of a hierarchical cel-lular mobile communication system,”ICUPC, pp. 128–132, 1993.

[4] T. P. Chu and S. S. Rappaport, “Overlapping coverage and channelrearrangement in microcellular communication systems,”IEE Proc.Commun., vol. 142, no. 5, pp. 323–332, Oct. 1995.

[5] G. L. Choudhury and S. S. Rappaport, “Cellular communicationschemes using generalized fixed channel assignment and collision typerequest channels,”IEEE Trans. Veh. Technol., vol. VT-31, pp. 53–65,May 1982.

[6] S. S. Rappaport and L. R. Hu, “Microcellular communication systemswith hierarchical microcell overlays: Traffic performance models andanalysis,”Proc. IEEE, vol. 82, pp. 1383–1397, Sept. 1994.

[7] , “Personal communication systems using multiple hierarchical cel-lular overlays,”IEEE J. Select. Areas Commun., vol. 13, pp. 406–415,May 1995.

[8] C. Chang, C. J. Chang, and K. R. Lo, “Analysis of a hierarchical cellularsystem with reneging and dropping for waiting new and handoff calls,”IEEE Trans. Veh. Technol., vol. 48, pp. 1080–1091, July 1999.

[9] R. Beraldi, S. Maramo, and C. Mastroianni, “A reversible architecturefor microcellular systems with overlaying macrocells,” inProc.INFOCOM’96, pp. 51–58.

[10] K. R. Lo, C. J. Chang, C. Chang, and B. Shung, “A combined channelassignment strategy for hierarchical cellular systems,” inProc. ICUPC,1997, pp. 128–132.

[11] K. L. Yeung and S. Nanda, “Channel management in microcell/macro-cell cellular radio systems,”IEEE Trans. Veh. Technol., vol. 45, pp.601–612, Nov. 1996.

[12] C. J. Chang, T. T. Su, and Y. Y. Chiang, “Analysis of a cutoff prioritycellular radio system with finite queueing and reneging/dropping,”IEEE/ACM Trans. Networking, vol. 2, pp. 166–175, Apr. 1994.

[13] M. Sugeno,Industrial Applications of Fuzzy Control. Amsterdam, TheNetherlands: North-Holland, 1985.

[14] T. Munakata and Y. Jani, “Fuzzy systems: An overview,”Commun.ACM, vol. 37, no. 3, pp. 69–76, Mar. 1994.

[15] L. A. Zadeh, “Fuzzy logic, neural networks, and soft computing,”Commun. ACM, vol. 37, no. 3, pp. 77–84, Mar. 1994.

[16] V. Catania, A. Puliafito, M. Russo, and L. Vita, “A VLSI fuzzy inferenceprocessor based on a discrete analog approach,”IEEE Trans. Fuzzy Syst.,vol. 2, May 1994.

[17] K. W. Shum and C. W. Sung, “Fuzzy layer selection method in hierar-chical cellular systems,” inProc. GLOBECOM’96, pp. 1049–1053.

[18] V. Catania, G. Ficili, S. Palazzo, and D. Panno, “A comparative Analysisof fuzzy versus conventional policing mechanisms for ATM networks,”IEEE/ACM Trans. Networking, vol. 4, pp. 449–459, June 1996.

[19] R. G. Cheng and C. J. Chang, “Design of a fuzzy traffic controllerfor ATM networks,” IEEE/ACM Trans. Networking, vol. 4, no. 3, pp.460–469, June 1996.

[20] M. Abdul-Haleem, K. F. Cheung, and J. Chuang, “Fuzzy logic baseddynamic channel assignment,”ICCS, pp. 773–777, 1994.

[21] H. T. Nguyen, M. Sugeno, R. Tong, and R. R. Yager,Theoretical Aspectsof Fuzzy Control. New York: Wiley, 1995.

[22] Q. Hu and W. Petr, “A predictive self-tuning fuzzy-logic feedback ratecontroller,” IEEE/ACM Trans. Networking , submitted for publication.

[23] M. Hellebrandt, R. Mathar, and M. Scheibenbogen, “Estimating positionand velocity of mobile in a cellular radio network,”IEEE Trans. Veh.Technol., vol. 46, pp. 65–71, Feb. 1997.

[24] K. R. Lo, C. J. Chang, C. Chang, and B. Shung, “A combined channelassignment strategy for hierarchical cellular systems,”ComputerCommun., vol. 21, no. 13, pp. 1143–1152, 1998.

Kuen-Rong Lo was born in Hsinchu, Taiwan, in1957. He received the B.S. degree in electronicsengineering from National Taiwan Institute ofTechnology, Taipei, Taiwan, in 1984, and the M.S.degree in computer science from National TsingHua University, Hsinchu, in 1989. He is currentlyworking toward the Ph.D. degree at National ChiaoTung University, Taiwan.

He is also a Project Leader in the Telecommunica-tion Laboratories Chunghwa Telecom Co., Ltd. Hisresearch interests include broad areas of mobile com-

munication systems and integrated services digital networks. Presently, his re-search is focused on channel assignment schemes, traffic performance mod-eling, and analysis for mobile cellular communication systems.

Chung-Ju Chang (S’81–M’85–SM’94) was bornin August, 1950. He received the B.E. and M.E.degrees in electronics engineering from the NationalChiao Tung University, Hsinchu, Taiwan, in 1972and 1976, respectively, and the Ph.D. degree inelectrical engineering from the National TaiwanUniversity, Taiwan, in 1985.

From July 1976 to August 1988, he was withthe Division of Switching System Technologies,Telecommunication Labs, Directorate General ofTelecommunications (DGT), Ministry of Commu-

nications, Taiwan, as a Design Engineer, Supervisor, Project Manager, andthen Division Director. There, he was involved in designing digital switchingsystem, ISDN user-network interface, and ISDN service and technology trials.In the meantime, he had acted as a Science and Technical Advisor for theMinistry of Communications from 1987 to 1989 and once helped DGT in theintroduction of digital switching systems in 1987. In August 1988, he joinedthe faculty of the Department of Communication Engineering, College ofElectrical Engineering and Computer Science, National Chiao Tung University,as an Associate Professor. He has been a Professor since 1993. He was Directorof the Institute of Communication Engineering from August 1993 to July 1995.His research interests include performance evaluation, wireless communica-tions networks, and broad-band networks. He had served as an Advisor for theMinistry of Education to promote the education of communication science andtechnologies for colleges and universities in Taiwan since 1995.

Dr. Chang is also acting as a Committee Member of the TelecommunicationDeliberate Body; a Committee Member of the Technical Review Assembly,Industrial Development Bureau, a Committee Member of the Electronics andInfomatics Development Committee, Ministry of Economic Affairs; a Com-mittee Member of the Telecommunication Administration Review Board, Min-istry of Transportation and Communications; and the Chairman of IEEE Vehic-ular Technology Society, Taipei Chapter. He is a member of CIE.

Cooper Changwas born in Keelung, Taiwan. He re-ceived the B.S. degree in the Department of Elec-trophysics from the National Chiao Tung University,Hsinchu, Taiwan, in 1990, and the M.E. degree inelectrical engineering from the National Tsing HuaUniversity, Taiwan, in 1992, and the Ph.D. degreefrom the Institute of Electron, National Chiao TungUniversity, Taiwan, in 1999.

He is a System Engineer with Acer Peripherals,Inc., Taiwan. His current research topics in API aresystem architecture design and performance evalua-

tion of wireless personal access network. His research interests include WPAN,embedded system, PCS, and 3G systems.


C. Bernard Shung received the B.S. degree in elec-trical engineering from National Taiwan Universityin 1981, and the M.S. and Ph.D. degrees in electricalengineering and computer science from University ofCalifornia, Berkeley, in 1985 and 1988, respectively.

He is currently a Design Manager at AllayerCommunications. He was a Visiting Scientistof IBM Research Division, Almaden ResearchCenter, in 1988 to 1990, and later a Research StaffMember 1998–1999. He was a Faculty Memberin the Department of Electronics Engineering,

National Chiao Tung University in Hsinchu, Taiwan, R.O.C. from 1990 to1997. From 1994 to 1995, he was a Staff Engineer at Qualcomm Inc. in SanDiego, CA, while taking a leave from NCTU. His research interests includeVLSI architectures and integrated circuits design for communications, signalprocessing, and networking. He has published more than 50 technical papersin various research areas.

Documents

A QoS-guaranteed fuzzy channel allocation controller for hierarchical cellular systems