Handover management for a PCS network

Abstract

A handover is required in wireless personal communications systems when a portable moves from one base station coverage area to another during the course of a conversation. In general, the handover should be completed while the portable is in the overlap reglon to be successful

This article considers several issues for handover management, handover detection, channel assignment, and radio link transfer.

OERPEL, HUGHES N E T W O R K SYSTEMS ATIONAL C H I A O

TUNG

n a personal communications services (PCS) system, it is important to support services when a subscriber moves from the coverage area of one base station to that of another. This process is referred to as small-scale mobility management or handover management. To illustrate the

handover behavior, let us consider an example in Fig. 1. This figure shows three radio base stations and their coverage areas (or cells). Even though cellular base station towers are shown, this discussion is general and applies to both high-tier cellular systems and low-tier pedestrian systems, and to both indoor and outdoor environments. A mobile phone or portable (represented

by a vehicle in the figure) is shown traversing these areas. Cov- erage areas are irregular because of buildings, trees, mountains, and other terrain features, and sometimes they may consider- ably overlap. A handover is required in mobile or portable communications systems when a portable moves from one base station coverage area to another during the course of a conversa- tion. In general, the handover should be completed while the portable is in the overlap region to be successful. As a portable is moved toward the edge of coverage of a base station, the sig- nal strength and quality begin to deteriorate. At some point the signal from a neighbor base station becomes stronger. Addi- tionally, the second base station receives a stronger signal from the portable than that received by the original base station. At some point the conversation needs to be handed over to the second base station before the link between the first base sta- tion and the portable becomes unusable and the call is lost.

Several issues need to be considered for handover manage- ment.

Handover Detection

To initiate a handover, two issues must be considered:

* Who initiates the handover process?

* How is the need for handover detected?

T h e decision on when to effect the handover must be based on measurements of the links made at the portable or

at the two base stations, or both.

While it is obvious that the measurements can be made at either the portable or the base stations, not so obvious is that the decision to effect the handover can be made by either the network or the portable. We will describe three strategies for handover detection.

Handovers are expensive t o execute, so needless han- dovers should be avoided. If the handover criteria are not chosen appropriately, t h e n , in t h e overlapping region between the two base station coverage area boundaries, the call might be handed back and forth several times between them. If the criteria are too conservative, the call might be lost before the handover can take place. The handover deci- sion-making criteria become even more critical with the evo- lution to smaller cell sizes, which is being done to increase the capacity of systems and to reduce the power require- ments of portable terminals. Unreliable and inefficient han- dover procedures will reduce the quality and reliability of the system. The propagation environment is dynamic and, even very close to the original base station, the received sig- nal at the portable could temporarily fade due to multipath propagation so that the signal from another base station might appear stronger for a brief period. During such brief fades it is not desirable to effect a handover since it would only be a temporary fix, and indeed the signal might return to normal much faster than the handover can be implement- ed. We will discuss several strategies that have been pro- posed to address these problems.

Channel Assignment

A channel assignment scheme must trade off the following performance objectives:

Implementation complexity

of

0 Spectrum utilization

The objective of a channel assignment strategy should be to achieve a high degree of spectrum utilization for a given grade of service with the least number of database lookups and the sim- plest algorithm employed in both the portable and the network.

Handover requests and initial access requests will compete for radio resources. At a busy base station, call attempts which fail because there are no available channels are called

blocked calls. Handover requests for existing calls which must

be turned down because there are no available channels are called forced terminations. It is generally believed that forced

terminations are less desirable than blocked call attempts. Several channel assignment strategies have been developed

which attempt to reduce forced ter- minations at the cost of increasing the number of lost or blocked calls. Several handover-initial access chan- nel assignment schemes, such as the

nonprioritized scheme, resewed chan- nel scheme, queuingpriority scheme,

and subrating scheme, have been pro-

posed. The reader is referred to for more details. Note that handover access success is intimately tied to the radio technology of the channel assignment process, which may be dynamic channel assignment (DCA), fixed channel assignment (FCA), quasi-static autonomous frequency assignment (QSAFA), or some other fixed, flexible, or dynamic process.

W Figure 1 . Handover.

Radio Link Transfer

There are two classes of link transfer procedures. For hard handover, the portable can only converse with one base station

at a time, and there is usually some small gap in the conversa- tion during the link transition. For soft handover [ 2 2 ] , the

portable can receive the signals from both base stations simul- taneously and both base stations can receive the portable. The network must combine the signals from the two base stations in some way. Soft handover, which is the more complicated of the two, is required of some code-division multiple access (CDMA) systems [12] because of the nearifar problem and the need for precise power control. As can be appreciated from Fig. 1, some calls will be in soft handover mode to more than two base stations. This article will focus on hard hand- over under time-division MA (TDMA) or frequency-division MA (FDMA) systems.

Two operations must take place for a successful handover: The air interface link must move from one base station to

another.

The network must bridge the second base station link into the existing call and drop the link to the first base station. We will discuss both operations.

Link transfer can be made from one channel to another on the same base station, or from one base station to another, both of which subtend the same controller or switch. In this case the “network” operation is relatively simple. Alternative- ly, the handover can take place between base stations whose common point is much higher in the switching hierarchy of the network, in which case the network operation can be expensive, time-consuming, and difficult.

We will provide detailed discussions and overviews of these handover issues.

Handover Detection

‘his section describes link measurement techniques to detect the need for handover, and discusses three strate-

T

Radio link Measurement

Handover detection is based on the link measurement pro- cess. The measurement process determines the need for hand- over and the target or new channel for transfer. T h r e e measurements are used to determine the quality of a channel:

WEI, or word error indicator, is an indication of whether the

current burst was demodulated properly in the portable. RSSI, or received signal strength indication, is a measure of

co-channel interference power and noise. The RSSI metric

has a large useful dynamic range (typically between 80 to 100 dB). QI, or quality indicator, is the esti-

mate of the “eye opening” of a radio signal, which relates to the signal-to-interference (S/I) plus noise ratio, including the effects of dispersion. Q I has a narrow range (relating to the range of S/I from 5 dB to perhaps 25 dB).

Handover may depend more reli- ably on WE1 (a compilation of the measured data for the desired signal over a period of time) of the current channel rather than RSSI (i.e., if WE1 is good, handover is not per- formed). However, it is required to accumulate WE1 measurements over

a period of time, whereas RSSI is known instantaneously. To make the handover decision accu- rately and quickly, it is desirable to use both WE1 and RSSI for the decision algorithm.

RSSI measurements are affected by distance-dependent fading, lognormal or shadow fading, and Rayleigh or multi- path fading. Ideally, the handover decision should be based on distance-dependent fading and, to some extent, on shadow fading, and should be independent of Rayleigh fading. This can be accomplished by averaging the received signal strength for a sufficient period of time. The problem is that besides transmitting and receiving the desired signals for the commu- nication link, the portable must also measure or sample all frequencies in the band of interest to find a suitable candidate for handover. Consider the TDMA system. Depending on the radio system’s TDMA frame structure and duration, it may take 100-500 ms to measure all of the possible frequency channels. Maintaining a short list of the best candidate chan- nels is a reasonable alternative since the number and frequen- cy of measurements of the most likely candidate base stations can be increased. Therefore, the decision will need to be based on a sum of instantaneous power measurements rather than a continuous measurement, which can thus average out the Rayleigh fading.

Channel comparisons for handover are based on the RSSI and Q I metrics. Since the multipath environment tends to make the RSSI and QI metrics vary widely in the short term, and since it is preferable not to perform handover to mitigate brief multipath fades (because these fades are nonreciprocal, and such handovers could cause unnecessary load on the net- work), the portable should average or filter these measure- ments before using them to make decisions.

The speed of the measurement process is dependent on the TDMA frame structure of the radio system. This capabili- ty can be used to visit each frequency channel in turn. The measurements obtained in this process are used to maintain an ordered list of channels as candidates for handover. The PACS radio system

,

ms. For a system with 25 frequency channels, this corresponds

to visiting each channel every 62.5 ms. A user moving at 1 m/s travels around 113 wavelength at 2 GHz in this time. If anten- na diversity is employed in the radio system at the portable, t h e greater of t h e two values would be qelected and the remaining measurement discarded.

Filtering should be applied to both RSSI and QI measure- ments. At least two methods of filtering are possible: window averaging and leab-bucket integration.

For window averaging, the portable maintains a number which is proportional to the average of the current mea-

surements and the last w - 1 measurements, where

w

sk = Sk-1 mk - mk-w,

where s k refers to the sum at time k , and mk to the mea- surement made at time period k. Note that the portable

must maintain a record of the current sample and the pre- vious samples.

* For leaky-bucket integration, the portable implements a dis- crete digital one-pole low-pass filter:

sk = ask-1

+

where a < 1 is a constant “forgetting factor.”

Either method is acceptable from a performance perspective. Note that handover should be initiated whenever the best channel has filtered RSSI (i.e., the “stable” RSSI value measured based on the techniques just mentioned) exceeding that of the current channel by some hysteresis value on the order of 6 dB.

A filtering process applied to the RSSI and QI metrics will reduce their usefulness in mitigating sudden (shadow) fades, for example, due to rounding a corner or slamming a door. The downlink WE1 can be used to detect and correct these “trouble” situations on an “override” basis. A count Cdown is kept of the number of downlink word errors (the count is reset every complete measurement cycle). If Cdown exceeds some threshold, the portable should initiate a handover if an appropriate channel can be found. Channel selection can fol- low the same process as given above; however, the hysteresis value can be lowered.

In order to reduce the potential tendency of a portable in certain circumstances to request a large number of handovers in quick succession, there should be a “dwell” timer, which keeps the portable from requesting a handover until some reasonable period of time after a successful handover. Refer- ence [20] discusses using the Doppler frequency to estimate the velocity of the vehicle and then the measurement averag- ing interval to average out both multipath and shadow fading. It thus affects handover only on the basis of path loss.

The propagation between the base station and the portable is made up of the direct line-of-sight (LOS) path and also scattering paths caused by reflections from or diffraction around buildings and terrain. Thus the signal received by the portable at any point consists of a large number of generally horizontally traveling uniform plane waves whose amplitudes, phases, and angles of arrival relative t o t h e direction of motion are random. These plane waves interfere and produce a varying field strength pattern with minima and maxima spaced on t h e order of a quarter-wavelength apart. T h e portable’s received signal fades rapidly and deeply as it moves through this interference pattern. By reciprocity, the base sta- tion receiver experiences the same phenomena as the portable due to the portable’s motion. The envelope process of this fast fading phenomena i s Rayleigh distributed if there is no strong LOS component; otherwise, it is Rician.

As the portable moves, different scatterers and terrain change the plane waves incident on the portable antenna. Therefore, superimposed on the rapid multipath fading are slow variations in the average field strength of the interfer- ence pattern due to these new reflection and diffraction paths. This slower fading phenomena is called shadow fading and has a lognormal distribution.

As the portable moves away from one base station toward another, t h e signals received from t h e first base station become weaker due to increased distance from the base sta- tion or path loss, and those received from the second base sta- tion become stronger. This very slow effect is often masked by

multipath Rayleigh fading and lognormal shadow fading. Short-term Rayleigh fading is usually handled in mobile sys- tem designs by diversity techniques such as frequency hop- ping, multiple receivers, or correlators with variable delay lines and antenna diversity, and signal processing techniques such as bit interleaving, convolutional coding, and equalizers.

The longer-term shadow fading is usually compensated for in the system link budget margins by increasing transmitter power and co-channel reuse distance. Slow fading can usually be tracked by power control of the portable device.

The path loss component of fading must be handled by handing off to the new base station when the signal from the old base station becomes unusable. Handovers in response to multipath or shadow fading will usually result in too many handovers. In addition, since it takes from 20 ms to several seconds to implement a handover, such a strategy, is not even an effective remedy for fast fading. However, the detection and measurement of fast fading can play an important role in the handover detection and decision process. This may be especially true when we consider handing off between high- and low-tier radio systems or between macro- and microcells of the same system. Such is the case when the portable is in a vehicle moving at highway speeds through microcells. In this case, even though the signal quality and strength from the low-tier base station may be momentarily better than that from the serving macrocell or high-tier base station, a han- dover might not be practical because the vehicle’s speed will move the communicating portable too rapidly through the coverage area of the low-tier base station. This can cause the network to perform too many handovers or cause these han- dovers to be required so rapidly that they become ineffective due to the delay in setting them up. Thus, it can be appreciat- ed that if the portable velocity can be detected, this would be

an aid to the handover detection and decision-making process.

Handover Detection Strategies

Three handover detection strategies have been proposed for PCS networks. The schemes whereby the portable controls the handover are called mobile-controlled handover (MCHO) in

the literature, and schemes whereby the network exercises control are called network-controlled handover (NCHO). There

is a third class, where the network controls the handover but the portable assists with measurements of the links. This is called mobile-assisted handovev (MAHO). The evolution of

mobile communications is toward more decentralization, implying that both the management and setup of handover procedures will be partially or fully entrusted to the portable. These three strategies will be described, as will air interface standards which implement them.

Mobile-ConfroZled Handover - MCHO is the most popular technique for low-tier radio systems and is employed by both the European Digital European Cordless Telecommunications

(DECT) and the North American Personal Access Communi- cations System (PACS) air interface protocols. In this method the portable continuously monitors the signal strength and quality from the accessed port and several handover candidate ports. When some handover criteria is met, the portable checks the “best” candidate port for an available traffic chan- nel and launches a handover request.

The combined control of automatic link transfer (ALT) (handover) and time slot transfer (TST) (handover between channels on the same base station) in the portable is consid- ered desirable in order to:

e Off-load this task from the network

e Ensure robustness of the radio link by allowing reconnection

Control both handover and handover between channels on thus preventing unhelpful simultaneous triggering of the two processes [6, 41. The control of handover in the portable is made possible by the portable’s capability to make quality measurements of the current and candidate channels. The con- trol of handover between channels on the same base station in the portable is made possible by passing uplink quality infor- mation (in the form of a word error indicator) back to the portable on the downlink. Quality maintenance processing, described schematically in Fig. 2, consists of four components:

Ongoing measurements and processing of measurement data which allows the portable to monitor quality

The trigger decision mechanism, whereby the portable uses the processed measurement data to determine that some action (handover or handover between channels on the same base station) is required

The choice of the new frequency carrier for handover or the new time slot for handover between channels on the same base station (a process closely allied with the trigger decision) Execution of the handover or handover between channels

on the same base station (e.g., via a signaling protocol between the portable and network equipment)

In other words, in a portable an ongoing measurement process examines radio link quality information [4, 51. When certain criteria are reached, the process indicates the need for a handover. It then selects a channel. Finally, the portable, in concert with the network, executes the handover. The avail- able link quality information is obtained through various means, and is “data-reduced” in such a way to provide a man- ageable amount of data, while retaining enough information to make good decisions about quality maintenance actions. As part of the demodulation process, the portable receiver gener- ally obtains two pieces of information: RSSI and QI.

Measurements of QI for the current channel are available to the portable once per frame as a result of the demodula- tion process. During the period of each TDMA frame where the portable is not receiving or transmitting information for the current call, the unit has adequate time to make a diversi- ty measurement (QI and RSSI for each antenna) of at least one additional channel. The downlink WE1 also is available to the portable. The base station can also feedback uplink WE1 to the portable. This information would only require 1 bit of the downlink stream per burst.

Finally, handover between channels on the same base sta- tion must also be handled in the same context. This is done to ensure that handover between channels on the same base sta- tion, which mitigates only the uplink situation, will not be per- formed when a handover could be used t o substantially improve both the uplink and downlink. In PACS, because of the use of time-division multiplexing (TDM) on the downlink, the use of the uplink word error feedback can indicate the need for a handover between channels on the same base sta- tion. Because DECT uses dynamic channel allocation, both the uplink and the downlink can be improved by a channel transfer to the same base station. The handover times for DECT have been reported to be as low as 100-500 ms. For PACS it is about 20-50 ms.

the same base station in the same place

Se’ect new _Execute

--+

ALT or TST

Network-Controlled Handover - NCHO is employed by the

low-tier CT-2 Plus [32, 131 and the high-tier Advanced Mobile Phone Service (AMPS) [l] cellular system. In this method, the port monitors t h e signal strength and quality from t h e portable, and when these deteriorate below some threshold, the network arranges for a handover to another port. The net- work asks all the surrounding ports to monitor the signal from the portable and report the measurement results back to the

I

;

Figure 2. Portable quality maintenance processing.

network. The network then chooses a new port for the han- dover and informs both the portable (through the old port) and the new port. The handover is then effected. In current analog cellular systems (AMPS, TACS, NMT, and NAMPS) and in the low-tier CT-2 and CT-2 Plus systems, the portable plays a passive role in the handover process. The base stations supervise the quality of all current connections by making measurements of RSSI. The mobile switching center (MSC) will command surrounding base stations to make measure- ments of these links occasionally. Based on these measure- ments, the MSC makes the decision of when and where to effect the handover. Because of the large volume of signaling traffic in the network needed to collect the information and the lack of adequate radio resources at base stations to make frequent measurements of neighboring links, handover execu- tion time is on the order of many seconds . Since measure- ments cannot be made very often, accuracy is reduced. T o reduce the signaling load in the network, neighboring base stations do not send measurement reports continuously back to the MSC; therefore, comparisons cannot be made before the actual RSSI is below a certain threshold.

Mobile-Assisted Handover - MAHO is a variant of NCHO where the network asks the portable to measure the signals from surrounding ports and report those measurements back to the old port so that the network can determine whether a handover is required and to which port. This handover strate- gy is employed by the high-tier GSM cellular standard but not by any candidate low-tier PCS radio system standards.

For MAHO, the handover process is more decentralized. Both the portable and the base station supervise the quality of the link (i.e., RSSI, WEI, and sometimes quality). RSSI mea- surements of neighboring base stations is done by the portable. In GSM the portable transmits the measurement results to the base station twice a second. The decision as to when and where to execute the handover is still made in the network (i.e., the base station and MSC). GSM handover execution time is approximately 1 s. In both MAHO and NCHO sys- tems, network signaling is required to inform the portable about the handover decision made in the network ( i.e., on which new channel to begin communicating), transmitted on the failing link. There is some probability that the link will fail before this information can be transmitted to the portable and the call will be forced to terminate.

For MCHO, NCHO, and MAHO, handover failure can occur for a number of reasons. Some of them are listed below:

There is no available channel on the selected base station. Handover is denied by the network, either for lack of

resources (e.g., no bridge or no suitable channel card), o r

because the portable has exceeded some limit on the number of handovers which may be attempted in some period of time. The network takes too long to set up the handover after the

handover has been initiated.

The target link fails in some way during the execution of handover.

In systems using dynamic channel allocation (DCA), such as D E C T , handovers can fail due to resource blocking [7].

Radio Link Transfer

e use t h e PACS system to illustrate t h e link transfer process. In the PACS archi- tecture (Fig. 3), the povtables (repre- sented by t h e pedestrians in t h e figure) or fixed accessed units commu-

nicate with the network through the

radio ports (RPs) by using an air inter- face (inteiiace A), which takes the sig- nal over the air and converts it to a

wire or fiber transmittable signal. A group of channels (i.e., time slots) are assigned to each RP. Several RPs are

connected to a radio port control unit Figure 3. The PACS architecture (RPCU) through interface P , which

separates the R P signal into logical

channels. Also, the signal for managing radio function at mul-

tiple RPs is separated logically from the call traffic. The RPCU provides management and control functions between the RP and the local exchange network. Several RPCUs are connected to a switch of the local exchange network. The

access manager (AM) residing in the RPCU allows the RPCU

to invoke integrated services digital network (ISDN) features.

The AM functions include radio-related service control func- tions (e.g., multiple RP management, trunking provision, and RP-to-RP link transfers) and non-radio-related service control functions (e.g., call control, switching, and routing).

Depending on the network elements involved in handover, the PACS architecture introduces at least five distinct link transfer cases t o consider. T h e link may b e transferred

between:

*Two time slots or channels in the same port

*Two ports on the same RPCU, or intra- RPCU

*Two ports connected to different RPCUs on the same switch or MSC, or inter- RPCU

*Two ports connected to different RPCUs

on different switches, or interswitch *Two ports connected to different RPCUs

homing to different AMs, or inter-AM The network implementation of han- dovers will vary for each of these types of handover. They will have different aspects of their control, and yet must be integrated into a unified control protocol because the portable does not know, a priori, which sort of handover will be caused by the selection of its “best” RP. In PACS, M C H O is assumed. When a handover is needed, a new radio channel is selected by t h e portable, and a handover request message

is transmitted by the portable to the new RP. The handover can also be stimulated by t h e network. I t is, however, still t h e responsibility of the portable to choose the best RP. In the case of a handover failure, the portable link quality maintenance pro- cess must decide what to do next. It may choose to initiate another handover to the “next best” channel, to simply stay on the old channel, to try again later, or to per- form another action appropriate for the sit- uation.

W e use i n t e r - R P C U handover as an

dure in PACS. As defined above, such a handover takes place between RPs connected to different RPCUs homing on the same switch and AM. The message flow for the handover is illustrated in Fig. 4.

To initiate handover, the portable temporarily suspends the voice con- versation by sending a link suspend message to the old R P (message 1, Fig. 4) and sends a handover request message through an idle time slot of the new radio channel to the network (message 2, Fig. 4). Upon receipt of the acknowledgment from the net- work (message 3, Fig. 4), the portable returns to the old assigned channel by sending a link resume message to the old RPCU (message 4, Fig. 4) and continues voice communication while t h e network prepares for the han- dover. Note that the handover request and handover acknowledgment are transmitted in the clear since the new RPCU does not have the cipher key for the session. Also note that han- dover reauest messapes may collide

Figure 5. Inter-BS handover message flow (MAHO).

with e a c i o t h e r or wzh initial access attempts. That is, while one portable is seizing a channel for a handover, some other portable may seize the same channel for a handover, call orig- ination, registration, answer, and so on. The result could be loss of both messages, or capture of one over the other.

Upon receipt of a handover request message, the RPCU checks to see if it already controls this call. If so (i.e., it is an intra-RPCU handover), it sends a handover complete message and reconfigures itself to effect the handover. If not (i.e., it is at least an inter-RPCU handover, a case studied in our exam- ple), the RPCU sends a handover acknowledgment (message 3, Fig. 4) and marks the slot busy. The network may check other parameters to ensure that it wishes to complete the handover. (For example, there may be a network management check to disallow handovers from portables which have been requesting too many of them.) If the network decides to pro- ceed with the handover process, it transfers the session priva- cy key to the privacy coder associated with the new channel, and the switch inserts a bridge into the conversation path and bridges in the new port. Finally, the network informs the portable to execute the handover via both the old and new channels by sending the handover execution messages (mes- sages 5 and 7, Fig. 4, are sent from the network to the portable via the old and the new RPs, respectively). The portable releases the old channel by sending an access release message to the old RPCU (message 6, Fig. 4). Note that messages 5

and 6 are not exchanged if the old channel fails before the new channel is established.

Once the portable has made the transfer to the new port it sends the network a handover complete message through the new channel (message 8, Fig. 4), and resumes voice communi- cation. The network can then remove the bridge from the path and free up resources associated with the old channel. Note that by the time the network bridge is complete the new R P C U has acquired t h e session key from either t h e old RPCU or t h e access manager depending on t h e network handover protocol. Thus, both messages 7 and 8 are transmit- ted in the cipher mode.

The use of the bridge eliminates the necessity for exact coordination of:

The switching of the path from the old RPCU to the new (by the network)

The transferring of the conversation from the old RP to the new (by the portable)

Bridges used for handover should be inserted as quickly as possible. Bridges in common use in existing switching systems (such as “loudest talker” and “additive” bridges) are believed to be adequate. However, it is possible that specific character- istics will be required in the future, possibly necessitating spe- cialized bridges. We note that the MCHO procedures for other PCS networks are similar to the PACS handover proce- dure just described. For example, DECT follows the similar MCHO procedure except that the selected new channel and the old channel may use the same carrier (frequency). The synchronization between the portable and the new base sta- tion is not necessary. Thus, the DECT handover is referred to as a seamless handover.

The network protocol for MAHO/NCHO and MCHO sys- tems are different. Figure 5 shows the message flow design for handover in the GSM system using MAHO. The portable transmits the old BS the radio link measurement. In GSM, this information is updated every 0.5 s. When the old BS determines that a handover is required, it sends a handover required message (message 2, Fig. 5 ) to the switch. Note that in this case the handover is originated by the old BS as com- pared to the MCHO case, where the handover is initiated by the new RPCU.

When the switch receives message 2, it culls the list of can- didate BSs supplied by the old BS, and selects the highest ranked BS with an available channel. Then it sends a han- dover request message (message 3, Fig. 5) to the new (i.e., selected) BS.

When the new BS acknowledges the request (message 4,

Fig. 5), t h e switch sends t h e handover command mewage

(message 5, Fig. 5 ) with the information regarding the new BS

and the RF channel to the old BS.

The old BS commands the portable to transfer the link to the new BS (message 6, Fig. 5 ) . The portable tunes to the new R F channel, establishes the channel to the new BS, and sends the handover completion message to the new BS (message 7, Fig. 5).

The new BS informs the switch of the handover comple- tion (message 8, Fig. 5). The switch then clears the link to the old BS (message 9, Fig. 5). The handover procedure is com- pleted when the old BS acknowledges the clear command (message 10, Fig. 5).

In MAHO or NCHO, the handover command (message 6, Fig. 5 ) to the portable is sent on the failing link. The han- dover procedure fails if the portable does not receive this message. In MCHO, the handover request message (message 2, Fig. 4) is sent by the portable to the new RPCU on the new, more reliable, link. Therefore, the success of the han- dover does not depend on any signaling message on the fail- ing link. A n o t h e r advantage of MCHO is that it is not necessary to transmit the measurement information on the air interface, thus reducing the signaling overhead required to maintain the call.

his article describes three major issues for PCS handover management. On the handover detection issue, we describe who initiates the handover process and how the need for handover is detected. Further reading onthis issue can be found in [3-6, 10, 16-20, 23, 24, 30, 33, 34, 371.

On the channel assignment issue, we describe several strategies to handle a handover call when no channel is avail- able in the new base station. Further reading onthis issue can be found in 115, 26-28, 35, 36, 381.

On the link transfer issue, we use inter-RPCU handover as an example to illustrate the link transfer procedure in PACS. Other types of link transfer procedure are intersystem hand- over and interswitch handover 111, 251. Further reading on this issue can be found in 12, 8, 9, 18, 311.

Acknowledgment

Li Fung Chang provided valuable comments to improve the quality of this article.

References

[ I ] ANSI/EIA/TIA, "Mobile Station-Land Station Compatibility Specification," Tech. rep. 553, 1989.

[2] Bellcore, "Generic Criteria for Version 0.1 Wireless Access Communica- tions Systems (WACS) and Supplement," Tech. rep. TR-INS-001313,

issue 1 , 1994.

[31 R. C. Bernhardt,"An Improved Algorithm for Portable Radio Access," [4] R. C. Bernhardt, "Performance Aspects o f Two-way Transmission in

Portable Radio Systems," /€€E VTC, 1989.

[51 R. C. Bernhardt, "User Access in Portable Radio Systems in a Co-Channel Interference Environment," /€E€ JSAC, vol. 7, no. 1, 1989, pp. 49-58. [61 R. C. Bernhardt, "Time-Slot Management in Digital Portable Radio Sys-

tems," /E€€ Trans. Vehic. Tech., vol. 40, no. 1, 1991.

[7] Bout e t al., "A Mathematical Model for Dynamic Channel Selection in Conformity w i t h the Digital European Cordless Telecommunications Standard," Proc. /€E€ PIMRC. Seot. 1993.

IEEE vrc, 1988, pp. I 63-69.

[SI J. C. I. Chuang, "Autonomous Adaptive Frequency Assignment of TDMA Portable Radio Systems," /E€€ Trans. Vehic. Tech., vol. 40, no. 3, Aug.

1991, PD. 627-35.

191 J C I Chuang, "Performance Issues and Algorithms for Dynamic Chan- nel Assignment," Proc /E€€ GLOBECOM '92, Orlando, FL, 1992, Dec

[IO] R H Clarke, "A Statistical Model of Mobile-Radio Reception," Be// Sys Tech I , vol 47, 1968

[I 1 I EIMIA, "Cellular Radio-Telecommunications Intersystem Operations," Tech rep IS-41, rev B, 1991

[ I 21 EIMIA, "Mobile Station-Base Station Compatibility Standard for Dual- Mode Wideband Spread Spectrum Cellular System, Tech rep IS-95, 1993

[I31 ETSI, "Common Air Interface Specification t o Be Used for the Inter- working between Cordless Telephone Apparatus in the Frequency Band

6-9

864.1 MHz, Including Public Access Services," I-ETS 300 131:1990, June

1991.

[I41 ETSI, Digital European Cordless Telephone.Common Interface," v.

05.03, 1991.

[I51 G. J. Foschini, B. Gopinath, and Z. Miljanic, "Channel Cost of Mobili- ty," /€€E Trans. Vehic. Tech., 42, vol. 4, Nov. 1993, pp. 414-24. [I61 M . J. Cans, "A Power-Spectral Theory of Propagation in the Mobile-

Radio Environment, /€€€ Trans. Vehic. Tech., VT-21, no. 1, 1972.

[ I 71 Gudmundson, "Analysis of Handover Algorithms," I€€€ VTC, 1991. [I81 B. Gudmundson and 0. Grimlund, "Handoff in Microcellular Based Per-

sonal Telephone Systems," WINLAB Wksp., 1990.

[I91 Hata and Nagatsu, "Mobile Location Using Signal Strength Measure- ments i n a Cellular System," /€€€ Trans. Vehic. Tech, VT-29, no. 2, 1980.

[201 Holtzman, "Adaptive Measurement Intervals for Handoffs," Tech. rep. WINLAB-TR-26, Rutgers Univ., 1992.

[21] D. Hong and S. S. Rappaport, "Traffic Model and Performance Analysis for Cellular Mobile Radio Telephone Systems w i t h Prioritized and No- Protection Handoff Procedure," /€€€ Trans. Vehic. Tech., VT-35, no. 3,

Aug. 1986, pp. 7-92.

[221 B. Jabbari et al., "Network Issues for Wireless Communications," /E€€ Commun. Mag., Jan. 1995.

E231 Kanai and Furuya, "A Handoff Control Process and Microcellular Sys- tems," IEEE VTC, 1988.

[241 W. C. Y. Lee, Mobile Ce//u/ar Telecommunications Systems, McGraw- Hill, 1989.

[251 Y.-B. Lin and S . K. DeVries, "PCS Network Signaling Using 557," /€€E

Pers. Commun., June 1995, pp. 44-55.

[261 Y:B. Lin, S. Mohan, and A, Noerpel, "Channel Assignment Strategies for Hand-off and Initial Access for a PCS Network," / € € E Pers. Com-

mun., vol. 1, no. 3, 1994, pp. 47-56.

[271 Y.-B. Lin, S. Mohan, and A. Noerpel, "Queueing Priority Channel Assignment Strategies for Handoff and Initial Access f o r a PCS Net- work," /€€€ Trans. Vehic. Tech., vol. 43, no. 3, 1994, pp. 704-12. [281 Y.-B. Lin, A. Noerpel, and D. Harasty, "The Sub-rating Channel Assign-

ment Strategy for PCS Hand-offs," /€€€ Trans. Vehic. Tech., vol. 45, no.

1, Feb. 1996.

[291 M. Mouly and M.-B. Pautet, The GSM System for Mobile Communica- tions, published by the authors, 1992.

[301 Murase, Symington, and Green, "Handover Criterion for Macro and Microcellular Systems," /€E€ VTC, 1991.

[311 S. Nanda, K. Chawla, and K. Budka, "CDPD over Shared AMPS Chan- nels: Interference Analysis," Proc. /E€€ PIMRC, 1995.

[32] Radio Advisory Board of Canada, CT2Plus Class 2: Specification for the Canadian Common Air Interface for Digital Cordless Telephony, Includ- ing Public Access Services, annex 1 , "Radio Standards Specification 130, 1992.

[331 Sampath and Holtzman, "Estimation of Maximum Doppler Frequency for Handoff Decisions," /€E€ WC, 1993.

E341 Sampath and Holtzman, "Adaptive Handoffs Through Estimation o f Fading Parameters," /€E€ ICC, 1994.

[351 S. Tekinary and B. Jabbari, "Handover Policies and Channel Assignment Strategies in Mobile Cellular Networks," / € € E Commun. Mag., vol. 29,

no. 11, 1991.

[361 S . Tekinary and B. Jabbari, "A Measurement Based Prioritization Scheme for Handovers in Cellular and Microcellular Networks," /€E€

ISAC, Oct. 1992, pp. 1343-50.

[371 Vijayan and Holtzman, "A Model for Analyzing Handoff Algorithms,"

/€€E Trans. Vehic. Tech., vol. 42, no. 3, 1993.

[381 C. H. Yoon and K. Un, "Performance of personal portable radio tele- phone systems with and without guard channels," /€€€ISAC, vol. 11,

no. 6, pp. 911-917, Aug. 1993.

Biographies

ANTHONY NOERPEL'S (anoerpel@mnsinc.com) biography was unavailable when this issue went to press

Yi-BING LIN [5'80-M'96-SM'96] (liny@csie.nctu.edu.tw) received his B.S.E.E.

degree from National Cheng Kung University in 1983, and his Ph.D. degree in computer science from the University of Washington in 1990. From 1990

t o 1995, he was w i t h the Applied Research Area at Bell Communications Research (Bellcore), Morristown, New Jersey. In 1995, he was appointed professor in the Department o f Computer Science and Information Engi- neering (CSIE), National Chiao Tung University (NCTU). In 1996, he was appointed deputy director of the Microelectronics and Information Systems Research Center, NCTU. Since 1997, he has been elected chairman of CSIE, NCTU. His current research interests include design and analysis of personal communications services networks, mobile computing, distributed simula- tion, and performance modeling.