Download Handoff in Cellular Systems: Algorithms, Prioritization, and Implementation and more Study notes Mobile Computing in PDF only on Docsity!
andoff is a process of trans- ferring a mobile station (MS) from one base station (BS) or channel to another. The channel change due to handoff may be through a time slot, frequency band, codeword, or combi- nation of these for time-division multiple access (TDMA), fre- quency-division multiple access (FDMA), code-division multiple access (CDMA), or a hybrid scheme, respectively [1]. This article includes four major topics.
Topic 1: Deployment Scenarios and Handoff — Topic 1 describes different system deployment scenarios and their constraints on the handoff procedure. Handoff algorithms with a specific set of parameters cannot perform uniformly well in different communication system deployment scenarios since these scenarios are characterized by specific environ- ments. Examples of different system structures include macro- cells, microcells, overlays, integrated cellular systems, integrated cordless and cellular systems, and integrated terres- trial and satellite systems. Note that these system structures are expected to coexist in future wireless communication sys- tems and warrant closer study.
Topic 2: Resource Management in Cellular Systems — Topic 2 views handoff and other resource management tasks and details handoff-related system performance improve- ment. Prioritizing handoff is one way to improve handoff- related system performance. Several handoff prioritization schemes (e.g., guard channels and queuing) are discussed. Handoff represents one of the radio resource management tasks carried out by cellular systems. Some other resource management functions include admission control , channel assignment , and power control. If some of the resource man- agement tasks are treated in an integral manner, better over- all performance can be obtained in a global sense by making appropriate trade-offs.
Topic 3: Implementation of Handoff — Topic 3 describes how handoff procedure is implemented. The decision making pro- cess of handoff may be centralized or decentralized (i.e., the handoff decision may be made at the MS, BS, or mobile
switching center, MSC). Different systems use different approaches to execute the process of handoff, and handoff protocols characterize these approaches.
Topic 4: Analysis of Handoff Algorithms — Three basic mechanisms have been used to evaluate the performance of handoff algorithms, and these mechanisms — the analytical approach, simulation approach, and emulation approach — are described in topic 4.
Cellular System
Deployment Scenarios
The radio propagation environment and related handoff chal- lenges are different in different cellular structures. A handoff algorithm with fixed parameters cannot perform well in differ- ent system environments. Specific characteristics of the com- munication systems should be taken into account while designing handoff algorithms. Several basic cellular structures (e.g., macrocells, microcells, and overlay systems) and special architectures (e.g., underlays, multichannel bandwidth sys- tems, and evolutionary architectures) are described next. Inte- grated cordless and cellular systems, integrated cellular systems, and integrated terrestrial and satellite systems are also described.
Macrocells
Macrocell radii are in several kilometers. Due to the low cell- crossing rate, centralized handoff is possible despite the large number of MSs the MSC has to manage. The signal quality in the uplink and downlink is approximately the same. The transi- tion region between the BSs is large; handoff schemes should allow some delay to avoid flip-flopping. However, the delay should be short enough to preserve the signal quality because the interference increases as the MS penetrates the new cell. This cell penetration is called cell dragging. Macrocells have relatively gentle path loss characteristics [2]. The averaging interval (i.e., the time period used to average the signal strength variations) should be long enough to get rid of fading fluctuations. First- and second-generation cellular systems provide wide-area coverage even in cities using macrocells [3]. Typically, a BS transceiver in a macrocell transmits high out- put power with the antenna mounted several meters high on a
26 1070-9916/98/$10.00 © 1998 IEEE IEEE Personal Communications • December 1998
H
Handoff in Cellular Systems
Nishith D. Tripathi, Nortel
Jeffrey H. Reed and Hugh F. VanLandingham
MPRG, Virginia Tech
Abstract Handoff is an essential element of cellular communications. Efficient handoff algorithms are a cost-effective way of enhancing the capacity and QoS of cellular systems. This article presents different aspects of handoff and discusses handoff related features of cellular systems. Several system deployment scenarios that dictate specific handoff requirements are illustrated. An account of handoff-related resource management tasks of cellular systems is given. Implementation of the handoff process is explained. Several mechanisms for evaluation of handoff-related system performance are described.
This work was sponsored by the MPRG Industry Affiliates Program at Virginia Tech.
tower to illuminate a large area. Figure 1 shows three clusters of seven cells in a macrocellular system. A cluster con- sists of a group of cells marked A through G.
Microcells
Some capacity improvement techniques (e.g., larger bandwidths, improved meth- ods for speech coding, channel coding, and modulation) will not be sufficient to satisfy the required service demand. The use of microcells is considered the sin- gle most effective means of increasing the capacity of cellular systems [3]. Microcells increase capacity, but radio resource management becomes more difficult. Microcells can be classified as one-, two-, or three- dimensional, depending on whether they are along a road or a highway, covering an area such as a number of adjacent roads, or located in multilevel buildings, respectively [4]. Microcells can be classified as hot spots (service areas with a higher traf- fic density or areas that are covered poorly), downtown clus- tered microcells (contiguous areas serving pedestrians and mobiles), and in-building 3-D cells (serving office buildings and pedestrians) [5]. Typically, a BS transceiver in a microcell transmits low out- put power with the antenna mounted at lamppost level (approximately 5 m above ground) [3]. The MS also transmits low power, which leads to longer battery life. Since BS antennas have lower heights compared to the surrounding buildings, RF signals propagate mostly along the streets [6–8]. The antenna may cover 100–200 m in each street direction, serving a few city blocks. This propagation environment has low time dispersion, which allows high data rates [9]. Microcells are more sensitive to the traffic and interference than macrocells due to short-term variations (e.g., traffic and interference variations), medium/long-term alterations (e.g., new buildings), and incremental growth of the radio network (e.g., new BSs) [10]. The number of handoffs per cell is increased by an order of magnitude, and the time available to make a handoff is decreased [11]. Using an umbrella cell is one way to reduce the handoff rate. Due to the increase in the microcell boundary cross- ings and expected high traffic loads, a higher degree of decentralization of the handoff process becomes necessary [1]. Microcells encounter a propagation phenomenon called the corner effect. The corner effect is characterized by a sud- den large drop (e.g., 20–30 dB) in signal strength (e.g., at 10–20 m distance) when a mobile turns around a corner. The cor- ner effect is due to the loss of the line of sight (LOS) component from the serving BS to the MS. The corner effect demands a faster handoff and can change the signal quality very fast. The corner effect is hard to predict. A long
measurement averaging interval is not desirable due to the corner effect. Mov- ing obstacles can temporarily hinder the path between a BS and an MS, which resembles the corner effect. In a microcellular system there may be two types of handoff scenarios: an LOS handoff and a non-LOS (NLOS) handoff. An LOS handoff is a handoff from one LOS BS to another LOS BS. An NLOS handoff is a handoff from an NLOS BS to an LOS BS. In an LOS handoff, premature handoff requests should be prevented. In an NLOS hand- off, the handoff must be done as fast as possible as the user turns the corner. Some of the solutions to deal with these different requirements for LOS and NLOS handoffs in micro- cells are umbrella cells, macrodiversity, and switching to mobile- controlled handoff [2]. Reference [12] studies the properties of symmetrical cell plans in a Manhattan-type environment. Cell plans affect sig- nal-to-interference ratio (SIR) performance in the uplink and downlink significantly. Symmetrical cell plans have four nearest co-channel BSs located at the same distance. Such cell plans can be classified into half-square (HS), full-square (FS), and rectangular (R) cell plans. These cell plans are described next.
Half-Square Cell Plan — This cell plan places BSs with omnidirectional antennas at each intersection, and each BS covers half a block in all four directions. This cell plan avoids the street corner effect and provides the highest capacity. This cell plan has only LOS handoffs. Figure 2 shows an example of a half-square cell plan in a microcellular system.
Full-Square Cell Plan — There is a BS with an omnidirectional antenna located at every other intersection, and each BS cov- ers a block in all four directions. It is possi- ble for an MS to experience the street corner effect for this cell plan. The FS cell plan can have LOS or NLOS handoffs. Figure 3 shows an example of a full- square cell plan in a microcellular system.
Rectangular Cell Plan — Each BS covers a fraction of either a horizontal or verti- cal street with the BS located in the mid- dle of the cell. This cell plan can easily be adapted to market penetration. Fewer BSs with high transmit power can be used initially. As user density increases, new BSs can be added with reduced transmit power from appropriate BSs. The street corner effect is possible for this cell plan. The R cell plan can have LOS or NLOS handoffs. Figure 4 shows an example of a rectangular cell plan in a microcellular system.
Macrocell/Microcell Overlays
Congestion of certain microcells, the lack of service of microcells in some areas, and high speed of some users are some
n Figure 1. Seven-cell clusters in a macro- cellular system.
G
F
A
B
E
C
D
C
F
G
D
B
D
C
A
E
E
A
B
G
F
n Figure 2. A half-square cell plan in a microcellular system.
Base station (^) Cell A
Cell B
n Figure 3. A full-square cell plan in a microcellular system.
Base station (^) Cell A
Cell B (^) Cell C
receive notch filters should be used at the micro- cell BSs. It shows that key parameters for such an overlay are the powers of the CDMA BS and MS transmitters relative to the macrocell BSs and the MSs served by the macrocells. Reference [25] studies spectrum management in an overlay system. A new cell selection method is proposed, which uses the history of microcell sojourn times. A procedure to deter- mine an optimum velocity threshold for the pro- posed method is also outlined. A systematic approach to optimal frequency spectrum man- agement is described.
Special Architectures
There are several special cellular architectures that try to improve spectral efficiency without a large increase in infrastructure costs. Some of these structures, discussed here, include an under- lay/overlay system (which is different from the overlay/underlay system described earlier) and a multichannel bandwidth system. Many cellular systems are expected to evolve from a macrocellular system to an overlay/underlay system. A study that focuses on such evolution is described in [26].
Underlay/Overlay System — An underlay/overlay system is different from the overlay/underlay system described earlier. In an overlay/underlay system, frequency spectrum is divided between the macrocells and microcells in such a way that a macrocell uses certain channels throughout the cellular system [27]. Also, the macrocell typically has a separate BS and a transmission tower. However, in an underlay/overlay system, a tighter reuse factor is used within an overlay. For example, assume that there are 36 channels in a clus- ter of 12 cells. If there is no overlay or underlay, three chan- nels will be available for each cell. In the conventional overlay/underlay system, two channels per cell can be used in a cluster of 12 microcells, while the macrocell will use the remaining 12 channels throughout the cluster region. If uni- form distribution of traffic is assumed, the effective number of channels per cell will still be three (two channels from a microcell and one from a macrocell). On the other hand, in one arrangement of an underlay/overlay scheme, two reuse factors, 12 and 6, will be used instead of just one reuse fac- tor 12, as shown in Fig. 6. Within a cluster of 12 cells, two channels per cell will be used in an overlay system (channels O 1 through O 24 in Fig. 6), and the remaining 12 channels will be distribut- ed using the reuse factor of six (chan- nels U 1 through U 12 in Fig. 6). Thus, within a single overlay cluster there will be two underlay clusters, and each underlay cluster has a reuse factor of 6. Hence, effectively there will be four channels per cell in an underlay/overlay system compared to three channels per cell for a non-underlay/overlay system. Further improvement in capacity can be obtained by using an even tighter reuse factor of 3 in an underlay cluster. In this case, there will be four underlay clusters within an overlay cluster. The overlay cluster uses two channels per cell, and the underlay cluster uses four channels per cell. Thus, effectively six channels per cell will be available. The underlay/overlay scheme can enhance
capacity of the system without the infrastructure costs because the same BSs, transmission towers, and other hard- ware can be shared.
A Multiple-Channel-Bandwidth System — Multiple channel bandwidths can be used within a cell to improve spectral effi- ciency. In a multiple-channel-bandwidth system (MCBS), a cell has two or three ring-shaped regions with different band- width channels [28]. Figure 7 shows an MCBS. Assume that 30 kHz is the normal bandwidth for a signal. Now, for a three-ring MCBS, 30 kHz channels can be used in the outermost ring, 15 kHz channels in the middle ring, and 7. kHz channels in the innermost ring. The areas of these rings can be determined based on the expected traffic conditions. Thus, instead of using 30 kHz channels throughout the cell, dif- ferent bandwidth channels (e.g., 15 kHz and 7.5 kHz) can be used to increase the number of channels in a cell. The MCBS uses the fact that a wide-bandwidth channel requires a lower carrier-to-interference ratio (C/I) than a narrow-bandwidth channel for the same voice quality. For example, C/I require- ments for 30 kHz, 15 kHz, and 7.5 kHz channel bandwidths are 18 dB, 24 dB, and 30 dB, respectively, based on subjective voice quality tests [28]. If the transmit power at a cell cite is the same for all the bandwidths, a wide channel can serve a large cell while a narrow channel can serve a relatively small cell. More- over, since a wide channel can tolerate a higher level of co-chan- nel interference (CCI), it can afford a smaller D/R ratio (the ratio of co-channel distance to cell radius). Thus, in the MCBS more channels become available due to multiple-bandwidth signals, and frequency can be reused more closely in a given service region due to different C/I requirements.
Integrated Wireless Systems
Integrated wireless systems are exempli- fied by integrated cordless and cellular systems, integrated cellular systems, and integrated terrestrial and satellite sys- tems. Such integrated systems combine the features of individual wireless sys- tems to achieve the goals of improved mobility and low cost.
Integrated Terrestrial Systems — Terres- trial intersystem handoff may be between
n Figure 6. An underlay/overlay system.
Underlay^ O cluster 1
Underlay cluster 2 Overlay cluster
O U1 U
O3 O U3 U
O5 O U5 U
O7 O U7 U
O9 O U9 U
O11 O U11 U
O13 O U1 U
O15 O U3 U
O17 O U5 U
O19 O U7 U
O21 O U9 U
O23 O U11 U
n Figure 7. A multiple-channel-band- width system.
7.5 kHz (^) 15 kHz 30 kHz
two cellular systems or between a cellular system and a cordless telephone system. Examples of systems that need intersystem handoffs include GSM–Digital European Cordless Telephone (DECT), CDMA in macrocells, and TDMA in microcells. When a call initiated in a cellular system controlled by an MSC enters a system controlled by another MSC, intersystem handoff is required to continue the call [29]. In this case one MSC makes a handoff request to another MSC to save the call. The MSCs need to have software for intersystem handoff if intersystem handoff is to be implemented. Compatibility between the concerned MSCs needs to be considered, too. There are several possible outcomes of an intersystem handoff [29]:
- A long-distance call becomes a local call when an MS becomes a roamer.
- A long-distance call becomes a local call when a roamer becomes a home mobile unit.
- A local call becomes a long distance call when a home mobile unit becomes a roamer.
- A local call becomes a long-distance call when a roamer becomes a home mobile unit. There is a growing trend toward service portability across dissimilar systems such as GSM and DECT [30]. For example, it is nice to have intersystem handoff between cordless and cel- lular coverage. Cost-effective handoff algorithms for such sce- narios represent a significant research area. This article outlines different approaches to achieving intersystem handoff. Simulation results are presented for handoff between GSM and DECT/Wide Access Communications System (WACS). The paper shows that a minor adjustment to the DECT speci- fication can greatly simplify the implementation of an MS capable of intersystem handoff between GSM and DECT.
Integrated Terrestrial and Satellite Systems — In an integrat- ed cellular/satellite system, the advantages of satellites and cellular systems can be combined. Satellites can provide wide- area coverage, completion of coverage, immediate service, and additional capacity (by handling overflow traffic). A cellu- lar system can provide a high-capacity economical system. Some of the issues involved in an integrated system are dis- cussed in [31]. In particular, the procedures of GSM are examined for their application to the integrated systems. The future public land mobile telecommunication system (FPLMTS) will provide a personal telephone system that enables a person with a handheld terminal to reach anywhere in the world [32]. The FPLMTS will include low Earth orbit (LEO) or geostationary Earth orbit (GEO) satellites as well as terrestrial cellular systems. When an MS is inside the cov- erage area of a terrestrial cellular system, the BS will act as a relay station and provide a link between the MS and the satel- lite. When an MS is outside the terrestrial system coverage area, it will have a direct communication link with the satel- lite. Different issues such as system architecture, call handling, performance analysis of the access, and transmission protocols are discussed in [32]. The two handoff scenarios in an inte- grated system are described below.
Handoff from the Land Mobile Satellite System to the Ter- restrial System — While operating, the MS monitors the satellite link and evaluates the link performance. The received signal strengths (RSSs) are averaged (e.g., over a 30 s time period) to minimize signal strength variations. If the RSS falls below a certain threshold N consecutive times (e.g., N = 3), the MS begins measuring RSS from the terrestrial cellular sys- tem. If the terrestrial signals are strong enough, handoff is made to the terrestrial system, provided that the terrestrial system can serve the MS.
Handoff from the Terrestrial System to the Land Mobile Satellite System — When an MS is getting service from the terrestrial system, the BS sends an acknowledge request (called page ) at predefined intervals to ensure that the MS is still inside the coverage area. If an acknowledge request signal from the MS (called page response ) is not received at the BS for N consecutive times, it is handed off to the land mobile satellite system (LMSS). Reference [33] focuses on personal communication systems with hierarchical overlays that incorporate terrestrial and satel- lite systems. The lowest level in the hierarchy is formed by microcells. Macrocells overlay microcells and form the middle level in the hierarchy. Satellite beams overlay macrocells and constitute the topmost hierarchy level. Two types of subscribers are considered, satellite-only and dual cellular/satellite. Call attempts from satellite-only subscribers are served by satellite systems, while call attempts from dual subscribers are first directed to the serving terrestrial systems with the satellites taking care of the overflow traffic. An analytical model for teletraffic performance is developed, and performance mea- sures such as traffic distribution, blocking probability, and forced termination probability are evaluated for low-speed and high-speed users.
Handoff Prioritization
One of the ways to reduce the handoff failure rate is to priori- tize handoff. Handoff algorithms that try to minimize the number of handoffs give poor performance in heavy traffic sit- uations [34]. In such situations, a significant handoff perfor- mance improvement can be obtained by prioritizing handoff.
Introduction to Handoff Priority
Channel assignment strategies with handoff prioritization have been proposed to reduce the probability of forced termi- nation [35, 36]. Two basic methods of handoff prioritization, guard channels and queuing, are explained next.
Guard Channels — Guard channels improve the probability of successful handoffs by reserving a fixed or dynamically adjustable number of channels exclusively for handoffs. For example, priority can be given to handoff by reserving N channels for handoffs among C channels in the cell [37]. The remaining ( C – N ) channels are shared by both new calls and handoff calls. A new call is blocked if the number of channels available is less than ( C – N ). Handoff fails if no channel is available in the candidate cell. However, this concept has the risk of underutilizing spectrum. An adaptive number of guard channels can help reduce this problem. Efficient usage of guard channels requires the determination of an optimum number of guard channels, knowledge of the traffic pattern of the area, and estimation of the channel occupancy time distributions.
Queuing of Handoff — Queuing is a way of delaying handoff [29]; the MSC queues the handoff requests instead of denying access if the candidate BS is busy. Queuing new calls results in increased handoff blocking probability. The probability of a successful handoff can be improved by queuing handoff requests at the cost of increased new call blocking probability and a decrease in the ratio of carried-to-admitted traffic since new calls are not assigned a channel until all the handoff requests in the queue are served. Queuing is possible due to the overlap region between the adjacent cells in which MS can communicate with more than one BS. If handoff requests occur uniformly, queuing is not need- ed; queuing is effective only when handoff requests arrive in
ning but must determine whether co-channel usage is allowed or not. If adaptation to the changing propagation and interfer- ence conditions is done in a channel allocation algorithm, such an algorithm must guarantee a safe co-channel reuse dis- tance. Hence, a measure of interference for handoff candidate channel is required as an input to the channel allocation algo- rithm. Reference [44] deals with DCA using an artificial neu- ral network (ANN). In microcells, the variations in the telephone traffic load are large compared to those in macro- cells. Reference [45] proposes a DCA algorithm that adapts to these variations for a one-dimensional cellular system. The proposed algorithm maximizes the number of assigned calls and is suitable for distributed implementation. DCA gives bet- ter performance than FCA at low loads since it can adapt to traffic bursts. However, at high loads, it does not perform as well. Hence, some hybrid schemes have been suggested. Flexible channel assignment (FLCA) distributes some chan- nels among the cells in a cluster permanently and keeps the remaining channels for any cell’s use when that cell’s perma- nent channels are inadequate to cope with high traffic demand. As explained in the previous section, the use of guard chan- nels exclusively for handoff requests results in under-utilization of the scarce channel resources. Reference [46] presents a channel allocation algorithm that follows a most critical first policy in which a free channel is assigned to the handoff request which would be the first to be cut off if no channel were available at that time. Simulation results indicate that this algorithm is effective in reducing handoff failures. Reference [47] describes signal strength based distributed channel assign- ment schemes for a one-dimensional cellular system.
Power Control — Power control helps increase battery life, reduce health hazards, and contain interference. One way to exercise power control is to use SIR as a criterion. In this case, MSs try to attain a target SIR through continuous power adjustments. If the minimum possible power that meets the required C/I constraint at the receiver is transmitted, spec- trum efficiency increases (compared to the case of uncon- trolled transmit power). Increasing the transmit power (to improve C/I for better transmission quality) does not neces- sarily meet the objective since other transmitters in the system may also increase their power levels to reduce the interfer- ence caused to them, thus increasing the global interference level. This phenomenon is called the party effect.
Handoff — One of the easy solutions to BS assignment is to assign the MS to the nearest BS. Intercell handoff can be viewed as an adaptive method of preserving the planned cell boundaries and subsequently reducing the interference. Adap- tation to the spatial distributions of radio traffic (or interfer- ence) can be done by modifying cell areas and shapes dynamically by adapting the handoff parameters. This effect is called cell breathing. In the directed retry method, if the best BS is not available, the second best BS is tried for handoff. However, directed retry increases the effectively used cell areas, increasing the global interference level.
Resource Management Integrated Handoff Algorithms
Some algorithms that combine two or more radio resource management tasks are described next.
Combined Intracell Handoff and Channel Assignment — Channel allocation algorithms that adapt to the instantaneous interference and traffic situation can lead to an easier plan- ning process. This is a tremendous advantage since the system grows stepwise with the traffic demand in most cases [10]. Reference [10] proposes an adaptive channel allocation
algorithm that is adaptive to traffic and interference. It assumes that C/I of the current channel is measured periodically. This algorithm consists of several steps, which are outlined below.
- For a new call setup or intercell handoff, reassignment is per- formed. In other cases, reassignment is performed if C/I old for the current channel is less than a threshold C/I check.
- Since it may not be feasible to calculate C/I for all the chan- nels, channels are checked until a channel with good C/I is found. This channel is taken as a candidate channel.
- If C/I of the candidate channel, C/I cand , and C/I old are less than a threshold, C/I block , the call is blocked.
- In the absence of call setup or intercell handoff, the candi- date channel is accepted only if C/I cand exceeds C/I old by some hysteresis value.
An Uplink SIR-Based Integrated Handoff Algorithm — Ref- erence [40] proposes an integrated resource management based on four SIR thresholds. The resource management tasks incorporated into the algorithm are admission control , power control , handoff , and channel allocation. A call is dropped when the SIR drops below γ drop ; for example, 16 dB for Advanced Mobile Phone System (AMPS). γ drop is consid- ered the minimum tolerable SIR for acceptable speech quali- ty. Power control is achieved by a target SIR threshold γ t. Each MS tries to attain γ t through power control. Call admis- sion control is achieved by an SIR threshold γ new. A new call attempt succeeds only if it can offer an SIR higher than γ new. This SIR threshold ensures that the system is not packed too tightly; otherwise, it may be difficult to find free channels for handoff. Moreover, a new call, if admitted, will not cause severe interference to existing calls. Handoff and channel assignment are combined in the sense that handoff is made to the minimum interference channel when SIR drops below γ ho.
An RSS-Based Integrated Handoff Algorithm — The algo- rithm proposed in [41] uses RSS and a transmission quality measure for the channels as handoff criteria. BS allocation, channel assignment, and power control are treated in an inte- gral manner. A new BS is selected in the case of a new call setup, and intercell handoff is based on signal strength and possibly some network criteria. The comparison between the candidate BSs is done under equal transmit power levels. Power control is performed to increase spectral efficiency.
An Integrated Power Control and Handoff Algorithm — Ref- erence [48] treats power control and BS assignment issues in an integral manner. The objective is to find a combination of BS assignment and transmit power to provide a feasible solu- tion to the minimum transmit power (MTP) problem. An algorithm called minimum power assignment (MPA) is pro- posed which iteratively solves the MTP problem. During an iteration of the algorithm, an MS chooses a combination of BS and transmit power for which minimum power is needed to maintain an acceptable C/I (assuming that the other MSs transmit fixed powers at the same time). Reference [49] also proposes a similar combined power con- trol and BS selection algorithm to achieve higher capacity in a spread-spectrum cellular system. The proposed algorithm adapts transmit powers of users and switches users between the BSs to minimize interference. The algorithm also reduces traffic conges- tion in a cell by moving the users to less congested adjacent cells.
Handoff Protocols
There are four basic types of handoff protocols: network-con- trolled handoff (NCHO), mobile-assisted handoff (MAHO), SHO, and mobile-controlled handoff (MCHO). As the hand-
off decision making process is decentralized (i.e., moving from NCHO to MCHO), handoff delay (i.e., the time required to execute a handoff request) decreases, but the measurement information available to make a handoff decision also decreas- es. These protocols are briefly described next.
Network-Controlled Handoff
In an NCHO protocol, the network makes a handoff decision based on measurements of the RSSs of the MS at a number of BSs. Sometimes the network sets up a bridge connection between the old and new BSs and thus minimizes the duration of handoff. In general, the handoff process (including data transmission, channel switching, and network switching) takes 100–200 ms and produces a noticeable click in the conversation. This click is imperceptible in a noisy voice channel; however, it is perceptible when handoff occurs at a reasonable signal quali- ty [50]. Information about the signal quality for all users is located at a single point (the MSC). This information facilitates resource allocation. According to [51], the overall delay can be of the order of 5–10 s. This type of handoff is not suitable for a rapidly changing environment and a high density of users due to the associated delay. NCHO is used in first-generation ana- log systems such as AMPS, Total Acceses Communications System (TACS), and Nordic Mobile Telephone (NMT) [50].
Mobile-Assisted Handoff
An MAHO protocol distributes the handoff decision process. The MS makes measurements, and the MSC makes decisions. According to [51], there can be a delay of 1 s; this delay may be too much to counteract the corner effect. In GSM, the BS subsystem (BSS) includes a base transceiv- er station (BTS) and a base station controller (BSC) [52]. The BTS is in contact with MSs through the radio interface and includes radio transmission and receiver devices and signal processing. The BSC is in contact with the network and is in charge of the radio interface management, mainly the alloca- tion and release of radio channels and handoff management. One BSC serves several BTSs, and several BSCs are connect- ed to one MSC. The handoff time (the time between handoff decision and execution) in GSM is approximately 1 s [3]. If the serving and target BTSs are located within the same BSS, the BSC for the BSS can perform handoff without the involve- ment of the MSC. This is referred to as intra-BSS handoff. When the MSC coordinates the handoff process, such handoff can further be classified as intra-MSC (within the same MSC) or inter-MSC (between MSCs) [53]. GSM-based handoff algo- rithms are evaluated in [19, 54–56]. An Interim Standard 95 (IS-95)-based system uses SHO in conjunction with MAHO. SHO is a “make before break” con- nection, that is, the connection to the old BS is not broken until a connection to the new BS is made. SHO utilizes the technique of macroscopic diversity. There are several varia- tions of SHO. The term soft handoff is used when old and new BSs belong to two different cells. The term softer handoff is used when the two signals correspond to the two different sec- tors of a sectorized cell [57]. When soft and softer handoffs occur simultaneously, the term soft-softer handoff is used. As far as the MS is concerned, there is no difference between SHO and softer handoff. For the network, additional hard- ware overhead is required for soft handoff. One channel ele- ment hardware and one BS-to-MSC trunk are required for each cell involved in SHO. Additional frame-by-frame selec- tion diversity is also required at the switch. No additional hardware is required at the BS for softer handoff since the channel hardware can be configured to transmit signal to mul- tiple sector antennas and use diversity combining techniques to process the signals from multiple sector antennas. The
handoff threshold needs to be small enough to bound the overall SHO percentage but large enough to allow efficient diversity combining. The MS needs more than one demodula- tor to exploit diversity combining techniques. SHO can increase the capacity if exercised carefully. SHO has an advantage of changing SIR distribution. The MSs far from a BS receive more signal energy, which reduces outage proba- bility. Another advantage of SHO is that increased signal energy reduces the switching of the call between the BSs. This reduces the computational load. In particular, proper selec- tion of the SHO region and its associated parameters can avoid the ping-pong effect common in hard handoff [58]. A disadvantage of SHO is that the mobile undergoing SHO occupies channels between different BSs and the switch (MSC). Moreover, SHO tends to increase the traffic in the wired channels in a fixed network. The greater the number of BSs involved in SHO, the more the traffic in the fixed net- work. SHO algorithms are focus of [57–59].
Mobile-Controlled Handoff
In MCHO the MS is completely in control of the handoff pro- cess. This type of handoff has a short reaction time (on the order of 0.1 s) and is suitable for microcellular systems [51]. The MS does not have information about the signal quality of other users, but handoff must not cause interference to other users. The MS measures the signal strengths from surrounding BSs and interference levels on all channels. A handoff can be initiat- ed if the signal strength of the serving BS is lower than that of another BS by a certain threshold. The MS requests the target BS for a channel with the lowest interference. MCHO is the highest degree of handoff decentralization. Some of the advantages of handoff decentralization are that handoff decisions can be made fast, and the MSC does not have to make handoff decisions for every mobile, which is a very diffi- cult task for the MSC of high-capacity microcellular systems [60]. MCHO is used in the European standard for cordless tele- phones, DECT [3]. The MS and BS monitor the current chan- nel, and the BS reports measurements — RSS and bit error rate (BER) — to the MS. The C/Is of free channels are also measured. The handoff decisions are made by the MS. Both intracell and intercell handoffs are possible. The handoff time is approximately 100 ms.
Handoff Evaluation Mechanisms
Three basic mechanisms used to evaluate the performance of handoff algorithms include the analytical, simulation, and emulation approaches. These mechanisms are described here.
The Analytical Approach
This approach can quickly give a preliminary idea about the performance of some handoff algorithms for simplified hand- off scenarios. This approach is valid only under specified con- straints (e.g., assumptions about the RSS profiles). Actual handoff procedures are quite complicated and are not memo- ryless. This makes the analytical approach less realistic. For real-world situations, this approach is complex and mathemat- ically intractable. Some of the analytical approaches appearing in the literature are briefly touched on below. The level crossings of the difference between the RSSs from two BSs were modeled as Poisson processes for station- ary signal strength measurements in [61, 62]. In [63], this ana- lytical work was extended to nonstationary signal strength measurements, and the level crossings were modeled as Pois- son processes with time-varying rate functions. The results in [61, 63] are useful for determining the averaging interval and hysteresis level to achieve an optimum balance between the
Earth model and a two-slope model. In the flat Earth model , one direct ray and another reflected ray (with 180˚ phase shift) contribute to the total received E-field. In reference [75], an empirical path loss two-slope model is suggested. The path loss increases with a certain slope to a threshold distance (called a breakpoint ) and then increases with a higher slope. In reality, wave propagation in microcells is complicated and consists of reflections and diffractions in addition to free space propagation. However, the main fea- tures of path loss can still be described by these empirical models. For certain parameter settings, the two-slope path loss model approaches the flat-earth model. For NLOS propagation , LOS propagation is assumed to the street corner. After the corner, propagation path loss is cal- culated by placing an imaginary transmitter at the corner with the transmit power equal to the power received at the corner from the LOS BS.
- The slow fading or large-scale fading model: According to [75], the distribution of the slow fading component is close to a log-normal distribution for a majority of LOS and NLOS streets with different standard deviations. The distri- bution is actually a truncated log-normally distributed varia- tion. In simulations, the variation should not be greater than ±3 σ. For the measurements obtained in reference [75], the average value of σ was found to be 4 dB for LOS streets and 3.5 dB for NLOS streets. Reference [77] pro- poses an exponential autocorrelation model for shadow fad- ing in mobile radio channels. The results show that the model fit is good for moderate and large cells; the predic- tions are less accurate for microcells due to multipath.
- The fast fading or small-scale fading model: Fast (or short term) fading is usually modeled as a Rician distribution where parameter K (Rice factor) varies with distance. When K = 0, the variation is Rayleigh fading. Reference [75] suggests a fast fading model in terms of polynomials based on the Rician distribution. Fast fading can usually be neglected since it gets averaged out due to a short correla- tion distance relative to that of shadow fading.
The Traffic Model — Traffic can be assumed to be uniform for macrocells. However, road structures need to be considered for microcells, and traffic can be allowed only along the streets. The new call arrival process is modeled as an independent Poisson process with a certain mean arrival rate. The new call durations are independent exponential random variables with a certain mean. In some simulation scenarios, the statistics of dwell time can be useful [78]. Dwell time is defined as the average time spent by an MS in a cell without handoff.
The Mobility Model — The MSs have different velocities fol- lowing a truncated Gaussian distribution.
Specific Simulation Models — A brief account of widely used simulation models is given here. References [42, 60, 61, 63, 79, 80] use a two-BS model that is simple and widely used for evaluating signal-strength-based algorithms. This model is suitable for small-size macrocells and LOS handoffs in microcells. In this model an MS travels from one BS to another in a straight line at a constant veloci- ty. The path loss is calculated using a single-slope formula, and shadow fading is assumed to be log-normal with an expo- nential correlation function. A model suitable for evaluating the performance of signal- strength algorithms is used in [55, 56]. The model has a four- cell neighborhood, and the MS travels from one BS to another in a straight line with constant velocity. The model assumes that there is no power control, and all BSs transmit at the
same power level. The path loss is calculated using Hata’s model, and shadow fading is log-normally distributed. Refer- ence [81] has a three-cell instead of the four-cell neighbor- hood in [55, 56]. Two routes of an MS in a cluster of seven cells are consid- ered in [54]. The first route is from one BS to another in which the MS crosses cell borders such that it is inside the overlapping region for a minimum duration of time. This route gives insight into the behavior of the handoff algorithm in the handoff area. The second route is from one BS to another in which the MS is in the overlapping region most of the time. This second route is more hostile than the first in terms of handoff complications. The four-cell model of [55] can easily be modified to create these two MS routes by adjusting the cell radii. Reference [82] uses an SIR-based model that can be used for integrated dynamic resource management tasks. Twenty BSs are uniformly spaced on a ring. The traffic model and mobility models used in [82] are the same as described earlier. The new calls are uniformly distributed throughout the ring. A model suitable for evaluating LOS and NLOS handoffs in a microcellular environment is used in [83]. The LOS and NLOS propagation models are similar to those described earlier. The log-normal shadow fading with exponential correlation function for slow fading and Rician fading model for fast fading are used. The model of [3] is suitable for a microcellular environ- ment. Two NLOS paths are considered which give insight into the behavior of handoff algorithms when there are multiple street crossings. The effect of C/I is studied in [3] for a partic- ular cell plan. A worst-case scenario (i.e., C/I of 12 dB) is used to account for interference. Reference [3] also studies the C/I distribution for the MS and BS. A comprehensive model for a microcellular system is pre- sented in [41]. This reference considers a Manhattan-like structure and places a BS at every other corner. At every street crossing, an MS either goes straight or turns with a given probability. The model is formed into a torus-like struc- ture to avoid edge effects. The LOS propagation model is taken from [74]. For the NLOS model, it is assumed that buildings are infinitely tall, and there is a fixed loss of 20 dB every diffraction street corner. Shadow fading is not consid- ered, but fast fading is modeled as Rayleigh fading. A comprehensive simulation model suitable for macrocel- lular and microcellular environments is described in [84–86]. The conventional macrocellular environment is modeled by a 49-cell toroidal structure that has seven-cell clusters with 1 km radius cells [87]. The microcellular system has half-square cells with 100 m block size. The simulation model for a micro- cell system considers both the transmission and traffic charac- teristics. Such combined analysis of transmission and traffic characteristics provides a more realistic scenario for perfor- mance evaluation of a cellular system. Reference [88] gives a brief account of the simulation model (called M2 simulation ) developed at AT&T; this model includes the effects of propa- gation, traffic, and system configuration. The model of [38] is suitable for evaluating handoff perfor- mance in a mixed-cell environment. An urban Manhattan-like environment is simulated in which a cluster consists of four microcells. Four clusters cover the service area with a macro- cell overlaying the microcells. User mobility has been modeled as Gaussian with the mean value varying with the distance from the starting position of the MS. A sharp linear velocity decrease is adopted before turning, and a linear increase has been considered after the corner until the previous velocity is restored. The path loss is calculated using the two-slope law. Second and fourth powers are used. The street corner is simu- lated by a 4 dB/m linear decrease from the street corner and
lasting up to 20 m. After that an NLOS propagation is assumed. Slow fading is simulated by uncorrelated log-normal distribution. New calls follow a Poisson model and are uni- formly distributed along the streets.
The Emulation Approach
The emulation approach uses a software simulator consisting of a handoff algorithm to process measured variables (e.g., RSS and BER). Actual propagation-measurements-based sim- ulation has the advantage of giving better insight into the behavior of the radio channels and more accurate data. The main disadvantages are that this approach requires periodic measurement efforts and is not suitable for comparison of dif- ferent handoff algorithms on the same platform. Reference [11] uses measured data in handoff simulation (the measured data was obtained by conducting 1700 MHz experiments in an urban environment in southern England). The path loss was found to follow a two-slope formula with different slopes for different locations. The short-term fading was found to be Rician with Rice factors varying from 10 to zero depending on the distance between the MS and BS. It was found that the optimal handoff threshold level was differ- ent for different sites [11]. Reference [89] introduces an indoor propagation simula- tor. The indoor simulator models trace 13 rays over a cross- corridor and exhibit good agreement with the experiments of 950 MHz propagation with multipath fading. Reference [90] describes an experimental digital cellular system that consists of a private branch exchange (PBX)- based MSC, three BSs, two MSs, and a radio channel simula- tor. Experimental results indicate that a handoff decision can be made within 1 s, and the handoff procedure works well under typical microcell propagation conditions.
Conclusion
Handoff is an integral component of cellular communications. Efficient handoff algorithms can enhance system capacity and service quality cost effectively. Different system deployment sce- narios present different constraints on handoff procedure. Hand- off algorithms with a specific set of parameters cannot perform uniformly well in different communication system deployment scenarios since these scenarios impose distinct restrictions and peculiar environments on the handoff process. Such system sce- narios are illustrated. These system structures are expected to coexist in future wireless communication systems and warrant substantial study. Handoff prioritization can improve handoff- related system performance. Two basic handoff prioritization schemes, guard channels and queuing, are discussed. Handoff represents one of the radio resource management tasks car- ried out by cellular systems. Some other resource manage- ment functions include admission control, channel assignment, and power control. If the resource management tasks are treated in an integral manner, better overall performance can be obtained to achieve global goals by making appropriate trade-offs. Such integrated resource management is discussed briefly. Different systems use different approaches to execute the process of handoff, and handoff protocols that character- ize these approaches are explained. Several mechanisms can be used to evaluate handoff-related system performance; three such mechanisms are described in detail.
Acknowledgments
The authors are grateful to Dr. Ravi Jain of Bellcore, Dr. Magnus Almgren of Ericsson, Dr. Sirin Tekinay of New Jersey Institute of Technology, and the referees of the article for their helpful suggestions.
References
[1] G. Liodakis and P. Stavroulakis, “A novel approach in handover initiation for microcellular systems,” Proc. 44th IEEE VTC , 1994, pp. 1820–23. [2] G. P. Pollini, “Trends in Handover Design,” IEEE Commun. Mag. , Mar. 1996, pp. 82–90. [3] O. Grimlund and B. Gudmundson, “Handoff strategies in microcellular systems,” Proc. 41st IEEE VTC , 1991, pp. 505–10. [4] A. Salmasi and K.S. Gilhousen, “On the system design aspects of CDMA applied to digital cellular and Personal Communication Network,” Proc. 41st IEEE VTC , 1991, pp. 57–62. [5] J. Shapira, “Microcell engineering in CDMA cellular networks,” IEEE Trans. Vehic. Tech. , vol. 43, no. 4, 1994, pp. 817–25. [6] T. Kanai and Y. Furuya, “A Handoff Control Process for Microcellular Systems,” Proc. 38th IEEE VTC , 1988, pp. 170–75. [7] D. C. Cox and D. O. Reudink, “Layout and control of high capacity sys- tems,” Microwave Mobile Communications , New York: Wiley, 1974, W.C. Jakes Jr., Ed., pp. 542–622. [8] R. Beck and F. W. Ho, “Evaluation and performance of field strength relat- ed handover strategies for microcellular systems,” 3rd Nordic Sem. Digital Land Mobile Radio Commun., Copenhagen, Denmark, Sept. 12–15, 1988. [9] A. J. Ransom, “Handoff Considerations in Microcellular Systems Plan- ning,” Proc. PIMRC, 1995, pp. 804–8. [10] R. Beck and H. Panzer, “Strategies for handover and dynamic channel allocation in micro-cellular mobile radio systems,” Proc. 39th IEEE VTC , 1989, pp. 178–85. [11] S. Chia and R. J. Warburton, “Handover criteria for a city microcellular radio systems,” Proc 40th IEEE VTC ,1990, pp. 276–81. [12] M. Gudmundson, “Cell planning in Manhattan environments,” Proc. 42nd IEEE VTC , Apr. 1992, pp. 435–38. [13] M. E. Anagnostou and G. C. Manos, “Handover related performance of mobile communication networks,” Proc. 44th IEEE VTC , 1994, pp. 111–14. [14] S. Chia, “The control of handover initiation in microcells,” Proc. 41st IEEE VTC , 1991, pp. 531–36. [15] M. Frullone, P. Grazioso and G. Rive, “On the optimum allotment of frequency resources in mixed cellular layouts,” IEICE Trans. Fundamen- tals of Elect., Commun. and Comp. Sci. , Dec. 1992. [16] C.-L. I, L. J. Greenstein, and R. D. Gitlin, “A microcell/macrocell cellular architecture for low- and high-mobility wireless users,” IEEE JSAC , vol. 11, no. 6, Aug. 1993, pp. 885–91. [17] H. Furukawa and Y. Akaiwa, “A microcell overlaid with umbrella cell system,” Proc. 44th IEEE VTC , 1994, pp. 1455–59. [18] J. Worsham and J. Avery, “A cellular band personal communication systems,” Proc. 2nd Univ. Pers. Commun. , 1993, pp. 254–57. [19] K. Ivanov and G. Spring, Mobile speed sensitive handover in a mixed cell environment, Proc. 45th IEEE VTC , 1995, pp. 892–96. [20] J. F. Whitehead, “Cellular spectrum efficiency via reuse planning,” Proc. 35th IEEE VTC , 1985, pp. 16–20. [21] J. Naslund et al ., “An evolution of GSM,” Proc. 44th IEEE VTC , 1994, pp. 348–52. [22] C. W. Sung and W. S. Wong, “User speed estimation and dynamic channel allocation in hierarchical cellular system,” Proc. 44th IEEE VTC , 1994, pp. 91–95. [23] L. B. Milstein et al ., “On the feasibility of a CDMA overlay for personal com- munications networks,” IEEE JSAC , vol. 10, no. 4, May 1992, pp. 655–68. [24] D. M. Grieco, “The capacity achievable with a broadband CDMA micro- cell underlay to an existing cellular macrosystem,” IEEE JSAC , vol. 12, no. 4, May 1994, pp. 744–50. [25] K. L. Yeung and S. Nanda, “Channel management in microcell/macro- cell cellular radio systems,” IEEE Trans. Vehic. Tech. , vol. 45, Nov. 1996, pp. 601–12. [26] G. L. Lyberopoulos, J. G. Markoulidakis, and M. E. Anagnostou, “The impact of evolutionary cell architectures on handover in future mobile telecommunication systems,” Proc. 44th IEEE VTC , 1994, pp. 120–24. [27] V. Wille, S. M. Irons, and A. J. King, “Capacity increase in cellular radio networks using existing base station sites,” Proc. IEE Colloq. Propaga- tion Aspects of Future Mobile Sys. , 1996, pp. 4/1–8. [28] W. C. Y. Lee, Mobile Communications Design Fundamentals , Wiley, 2nd ed., 1993. [29] W. C. Y. Lee, Mobile Cellular Telecommunications , McGraw-Hill, 2nd ed., 1995. [30] N. W. Whinnett, “Handoff between dissimilar systems: general approaches and air interface issues for TDMA systems,” Proc. 45th IEEE VTC , 1995, pp. 953–57. [31] E. D. Re and P. Iannucci, “The GSM Procedures in an intergrated cellu- lar/satellite system,” IEEE JSAC , vol. 13, no. 2, Feb. 1995, pp. 421–30. [32] H. K. Lau, P. C. K. Liu, and K. C. Li, “Handoff analysis for an integrated satellite and terrestrial mobile switch over a fading channel,” Proc. IEEE PIMRC , 1992, pp. 397–401. [33] L.-R. Hu and S. S. Rappaport, “Personal communications systems using hierarchical cellular overlays,” Proc. IEEE ICUPC , 1994, pp. 397–401. [34] G. Senarath and David Everitt, “Performance of handover priority and queu- ing systems under different handover request strategies for microcellular mobile communication systems,” Proc. 45th IEEE VTC , 1995, pp. 897–901.
