当前位置:首页 >> >>

Radio resource management in future wireless networks requirements and limitations


Radio Resource Management in Future Wireless Networks – Requirements and Limitations
Jens Zander
Radio Communication Systems, Dept of Signals, Sensors & Systems Royal Institute of Technology, S-100 44 STOCKHOLM

Abstract
Comparing market estimates for wireless personal communication and considering recent proposals for wideband multimedia services with the existing spectrum allocations for these types of systems show that spectrum resource management remains an important topic in the near and distant future. In this paper, we start out by presenting a quite general formulation of the radio resource management problem where the three key allocation decisions are concerned with waveforms(“channels”), access ports (or base stations) and, finally, with the transmitter powers. We briefly review some the current approaches to these problems as found in the litterature. In particular the principles of random channel allocation schemes, as found in frequency hopping or direct sequence CDMA systems, are compared with determinisic dynamic channel allocation schemes. The paper closes by giving an outlook over some of the key problems in resource management in future wireless multimedia systems.

1. Introduction
The rapid increase of the size of wireless mobile community and their demands for high speed, multimedia communications stands in clear contrast to the rather limited spectrum resource that have been allocated in international agreements. Efficient spectrum or Radio Resource Management (RRM) is of paramount importance due to these increasing demands. Fig 1 illustrates the principles of wireless network design. The network consists of a fixed network part and a wireless access system. The fixed network provides connections between base stations or Radio Access Ports(RAP), which in turn provide the wireless "connections" to the mobiles. The RAPs are distributed over the geographical area where we wish to provide the mobile users with

ZANDER, RADIO RESOURCE MANAGEMENT

Fig 1. Wireless area communication system

communication services. We will refer to this area as simply as the service area. The area around a RAP where the transmission conditions are favourable enough to maintain a connection of the required quality between a mobile and the RAP, is denoted the coverage area of the RAP. The transmission quality and thus the shape of these regions will, as we may expect depend heavily on the propagation conditions and the current interference from other users in the system. The coverage areas are therefore usually of highly irregular shape. The fraction of the service area where communication with some required quality of service (QoS) is possible is called the coverage or the area availability of the system. In two-way communication systems(such as mobile telephone systems), links have to be establish both from the RAP to the mobile (down- or forward link) and between the mobile terminal and the RAP (up- or reverse link). At first glance these two links seem to have very similar properties, but there are some definite differences from a radio communication perspective. The propagation situation is quite different, in particular in wide area cellular phone systems, where the RAP (base station) usually has its antennas at some elevated location, free of obstacles. The terminals on the other hand are usually located amidst buildings and other obstacles creating shadowing and multipath reflections. Also the interference situation in the up and downlink will be different since there are many terminals and varying locations and quite few RAPs at fixed locations.

2(20)

ZANDER, RADIO RESOURCE MANAGEMENT

B1 M1 BB

M2

MM M3

Fig 2. Resource management problem formulation For obvious economical reasons, we would like our wireless network to provide ample coverage with as few RAPs as possible. Clearly this would not only minimise the cost of the RAP hardware and installation, but also limit the extent of the fixed wired part of the infrastructure. Coverage problems due to various propagation effects puts a lower limit to the number of RAP that are required. However not quite correct, one could say that the range of the RAPs is to small, compared to the inter-RAP distance. Such a system where this type of problem is dominant is called a range limited system. As the number of transmitters in the system becomes large within some fixed chunk of available RF-spectrum, the number of simultaneous connections (links) will become larger that the number of orthogonal signals that the available bandwidth may provide. In order to provide service for such a large population of users, it is obvious that the bandwidth used by the RAPs and terminals has to be reused in some clever way at the cost of mutual interference. The systems is said to be bandwidth or interference- limited . Absolutely vital to the study of any resource management problem is a thorough understanding of the user requirements, i.e. the required QoS and the traffic characteristics. All resource management schemes are designed (or optimised) using some model for the traffic. The resulting performance will clearly be a function of not only how well our design has been adapted to the traffic model, but also how accurate the traffic model is. Most wireless systems of today use circuit switched speech as the main design model (e.g. GSM). This does not prevent such systems to carry other types of traffic, but they always do this at a performance penalty. Future wireless access systems are expected to carry both large bandwidths as well as a mixture of services with very different and often conflicting service requirements. In particular in these scenarios, accurate modelling is

3(20)

ZANDER, RADIO RESOURCE MANAGEMENT

Bk Bi Gij Gil Mj Ml G kj G kl

Fig 3. Link Gains

imperative for efficient resource utilization. Unfortunately however, future user applications are little known and most work to derive realistic traffic models for these kind of applications lies still ahead of the telecommunication community. In the remainder of the paper we will now present a more rigorous formulation of the radio resource management and review some of the ideas and results from the literature. Finally wegive an outlook on how these results can be applied to future wideband systems and which are the key problems that should be addressed in further studies.

2. Radio resource management - a general problem formulation
Assume that M mobiles (M1, M1.... MM) are served by access ports (base stations), numbered from the set
B

= {1,2,3,...B}.

Now, let us assume that there are C (pairs of) waveforms (in conventional schemes these can be seen as orthogonal channels (channel pairs)) numbered from the set
C

= {1,2,3,...C}

available for establishing links between access ports and mobile terminals. To establish radio links, to each mobile the system has to assign

4(20)

ZANDER, RADIO RESOURCE MANAGEMENT

a) b) c)

an access port from the set B, a waveform (channel) from the set
C

,

a transmitter power for the access port and the terminal.

This assignment (of access port, channel and power) is performed according to the resource allocation algorithm (RAA) of the wireless communication system. The assignment is restricted by the interference caused by the access ports and mobiles as soon as they are assigned a "channel" and when they start using it. Another common restriction is that access ports are in many case restricted to use only a certain subset of the available channels. Good allocation schemes will aim at assigning links with adequate SIR to as many (possibly all) mobiles as possible. Note that the RAA may well (should) opt for not assigning a channel to an active mobile if this assignment would cause excessive interference to other mobiles. Let us now study the interference constraints on resource allocations in somewhat more detail. We now may compute the signal and interference power levels in all access ports and mobiles, given the link (power) gains, Gij, between access port i and mobile terminal j. For the sake of simplicity we will here consider only rather wideband modulation schemes which will make the link gains virtually independent of the frequency. Collecting all link gains in matrix form, we get a BxM rectangular matrix the link gain matrix G . The link gain matrix describes the (instantaneous) propagation conditions in the system. Note that in a mobile system, both the individual elements of the matrix (due to mobile motion) and the dimension of the matrix (due to the traffic pattern) may vary over time. The task of the Resource allocation scheme is to find assignments for the Quality-of-Service is sufficient in as many links as possible(preferably all). Providing a stringent definition to the QoS for a practical communication service is a complex and multifaceted problem. In this treatment we will confine ourselves to a simple maeasure the Signal-to-Interference ratio, SIR. or actually, to be precise, the Signal-to-interference+noise ratio. This measure is strongly connected with performance measures as the bit or message error probability in the communication link. We require the SIR to exceed a given threshold γo which is determined by the modulation and coding formats of the system. This means that the following inequality must hold for both the up (mobile-to-access port) and down (access port-to-mobile) link of the connection:

5(20)

ZANDER, RADIO RESOURCE MANAGEMENT

Γi =

PjGij

∑PmGimθjm+N
m

≥ γo

(1)

where Γj denotes the SIR at the receiver and N denotes the receiver (thermal) noise power at the access port. Pj denotes the transmitter power used by terminal j. The quantity θjm is the normalized crosscorrelation between the signals from mobiles j and m at the access port receiver, i.e. the effective fraction of the received signal power from transmitter m contributes to the interference when receiving the signal from access port j. If the waveforms are chosen to be orthogonal (as in FDMA and TDMA) these correlations are either zero or one depending on if the station has been assigned the same frequency (time slot) or not. In non-orthogonal access schemes (e.g.DS-CDMA) the θjm take real values between zero and one. Note that we may not be certain that it is possible to comply with all the constraints (2) for all the M mobiles, in particular if M happens to be a large number. As system designer we may have to settle for finding resource allocation schemes that assign channels with adequate quality to as many mobiles as possible. The largest number of users that may be handled by the systems is a measure of the system capacity. Since the number of mobiles is random quantity and the constraints (2) depend on the link matrix, i.e. on the relative position of the mobiles, such a capacity measure is not a well defined quantity. The classical approach for telephone type of traffic is to use as capacity measure the maximal relative arrival rate of calls ρ for which the blocking probability (the probability that a newly arrived session request is denied) can be kept below some predetermined level. Due to the mobility of the mobiles this is not an entirely satisfying measure. A call or session may be lost due to adverse propagation conditions. To include such phenomena in to our capacity would require detailed specification of call handling procedures (e.g. handling of new vs. old calls, hand-off procedures as a mobile moves from one access port to another etc.). It may therefore be practical to choose a simpler and more fundamental capacity measure that will reflect the performance of the resource allocation scheme as such. For this purpose, the assignment failure probability ν (or assigment failure rate [14,15]) has been proposed. The instantaneous capacity ω? (νo) of a wireless system is the maximum allowed traffic load in order to keep the assignment failure rate below some threshold level νo, i.e..

6(20)

ZANDER, RADIO RESOURCE MANAGEMENT

ω?(νo) = { max ω : ν ≤ νo }

(2)

As we have seen above, finding the optimum resource allocation, i.e. for each mobile determining i) a waveform assignment (determining theθjm) ii) an access port assigment (of one or more(!) ports) iii) a transmitter power assignment that maximises Y for a given link gain matrix, is a formidable problem. No efficient general algorithm that is capable of doing such an optimal assignment for arbitrary link gain matrices and mobile sets is known. Instead, partial solution and a number of more less complex heuristic schemes have been proposed (and are used in the wireless systems of today). These schemes are usually characterised by low complexity and by using simple heuristic design rules. The capacity

ω? achieved by these schemes is, as expected, often considerably lower that can be expected to be
achieved by optimum channel assignment.

3. Current Approaches to resource allocation strategies
The subproblem that has attracted most of the interest in the literature so far, is the choice and allocation of waveforms. Orthogonal waveforms such as frequency division multiplexing (FDMA) and time division (TDMA) that provide a "channelisation" of the spectrum have no doubt been the most popular ones, although considerable interest has recently been devoted to non-orthogonal waveforms, e.g. the IS-95 DS-CDMA waveforms[25]. Given the set of signalling waveforms
C

,

the next problem is the allocation of waveforms to the different terminal-access port links. This allocation can be done numerous ways depending on the amount and quality of the information available regarding the matrix G and the traffic situation (activity of different terminals). Another important issue is the time scale on which resource (re-)allocation is feasible. Channel allocation in early FDMA cellular radio systems operates on a long term basis. Based on average type statistical information regarding G (i.e. large scale propagation predictions), frequencies are on a more or less permanent basis assigned to different access ports. Such a "cell

7(20)

ZANDER, RADIO RESOURCE MANAGEMENT

plan" provides a sufficient reuse distance between RAPs providing a reasonably low probability of outage (to low SIR)[1]. Inhomogenuities in the traffic load can also be taken care of by adapting the number of channels in each RAP to the expected traffic carried by that access port. To minimise the planning effort, adaptive cell planning strategies (e.g. "channel segregation"[2]) have been devised using long-term average measurements of the interference and traffic to automatically allocate channels to the access port. These "static" (or "quasi-static") channel allocation schemes work quite well when employed in macrocellular systems with high traffic loads. In short range (microcellular) systems propagation conditions tend to change more abrubtly. Since each of the RAPs tend to carry less total traffic in small microcells, the relative traffic variation are also large, in particular in multimedia traffic scenarios. Employing "static" channel allocation schemes in the situations require considerable design margins. Large path loss variations are countered with large reuse distances, unfortunately at a substantial capacity penalty. In the same way microcellular traffic variations are handled by assigning excess capacity to handle traffic peaks. In recent years two principally different methods to approach this problem have been devised: Dynamic channel allocation (DCA) and Random Channel Allocation (RCA). In dynamic (Real-time) channel allocation(DCA), real-time measurements of propagation and/or traffic conditions are used to (re-)allocate spectrum resources. Early, graph theoretic schemes, adapted only to traffic variations[4,5]] yielding only moderate capacity gains (<50 %) compared to static systems in microcellular environments. Other schemes adapt their channel allocation to the received wanted signal strength. One example of the latter type of schemes is the class of Reusepartitioning schemes[6,8]. Here, several overlaid cell plans with different reuse distances are used. Terminals with a high received signal level are tolerant to interference and can be allocated a channel from a dense reuse cell plan whereas the "weaker" terminals get channels with a large reuse distance and lower interference levels. Capacity gains in the order of up to 100% have been reported for these schemes[7]. Also schemes directly estimating the C/I and thereby in a distributed way finding channels with adequate quality have been proposed[2,9]. Similar gains as in the reuse partitioning schemes are found in the literature. A comprehensive survey of different DCA-schemes is provided in [46]. The performance of the DCA schemes is critically dependent on the rate at which allocation or reallocation occurs. Purely traffic adaptive schemes act on incoming user request and users releasing capacity. Channel reallocation have to occur at these rates to fully utilise the potential of such a DCA scheme. For speech traffic this means that reallocations typically occur at second rates.

8(20)

ZANDER, RADIO RESOURCE MANAGEMENT

Γ
T1 T2 T4 γo T5

T3

t

a)

Γ
T3 T 1 T2 T5 T4 γo

t

b)

Fig 4. Typical realisations of terminal SIR:s in cellular systems with slowly moving terminals (a) and with rapidly moving terminals(b).

Path loss and interference adaptive schemes will have to "track" (at least slow fading) signal level variations and reallocation rates in the 10's of millisecond range may be required. An alternative class of allocation schemes are the random channel allocation schemes. The principle is most easily explained using fig 4. The graph 4a) shows a typical set of C/I-trajectories of five terminals in a cellular system. As we can see, 4 of the 5 terminals achieve an adequate C/I, corresponding to an (ensemble) outage rate of 20%. Compare this situation to the one in figure 4b) exhibiting the same outage rate. In contrast to the situation in a) where 20% of the terminals are

9(20)

ZANDER, RADIO RESOURCE MANAGEMENT

Ideal C/I balancing Pr[C/I<x] RSS PC No PC

Target C/I
Fig 5. Outage probability minimization - CDF of received C/I

x(db)

experiencing a to low C/I, here each terminal will experience insufficient quality 20% of the time. In case a) channel coding is a waste of capacity since 4 terminals have sufficient quality and the last unlucky terminal is probably "beyond salvage". In case b) however, there are probably a sufficient number of reliable channel symbols in all terminals to make reception possible, provided suitable constraint lengths & interleaving is used.The obvious way to achieve the latter situation irregardless of the mobile speed, is to permute channel allocations in a random fashion. The simplest way is to use (orthogonal) frequency hopping which can be seen as a static channel allocation where terminals allocated to a certain access port swap channels with each other[28,33]. Frequency hopping occurs typically 100-1000 times/second. Also non-orthogonal waveforms can be used as in the DS-CDMA based IS-95 scheme[25]. Effectively, a new random waveform is used for every transmitted bit. DS-CDMA schemes require only very low level of synchronisation and no cell planning which has made them attractive. Regarding capacity the comparison between DS and FH schemes is not obvious although orthogonal schemes seem to have advantages in mixed cell environments[26]. Comparing the performance of (deterministic) DCA to the performance of the Random allocation schemes is even more complex and stands out as one of the more fundamental research topics of the near future.

10(20)

ZANDER, RADIO RESOURCE MANAGEMENT

Another quite different situation where a similar interference conditions as in frequency hopping prevail are certain packet communication systems. Here the “randomness” is mainly induced by the random arrivals of packets triggering transmission events. The selection of the proper transmitter power in terminals and access port is another topic that has attracted considerable interest in recent years. There can be several objectives for this: to suppress adjacent channel (cross correlation) interference in non-orthogonal schemes, to minimise power consumption in order to extend terminal battery life and to control cochannel interference (in schemes with orthogonal waveforms). In the resource allocation problem context, it can be shown that the maximum number terminals is supported under a power control (PC) regime that balances the C/I of all terminals that can be supported and shuts of the rest[10]. Fig 5, showing the Cumulative Distribution Function (CDF) of the received C/I in an (orthogonal signalling) cellular system under three power control regimes, illustrates why this is so. As we can see, the uncontrolled system exhibits a rather flat CDF with a high outage probability (at the threshold C/I). A received signal strength based algorithm, e.g. the constant received power scheme, reduces some of the variations in the C/I by limiting the variations in the “C” component. The variations in the interference part (“I”) are however now larger than before and the figure shows typical net result CDF. The outage probability is now sligthly lower. In the C/I balancing scheme all stations have the same C/I, here slightly over the threshold, leaving only a small fraction of station without support. Finding this optimum set of non-supported terminals is a problem closely related to the design of DCA schemes. Distributed implementations and different implementational constraints[11,12] have been studied. Results show that very robust nearoptimum power control schemes can be devised at very low complexity. Performance result indicate that in static channel allocations substantial (>100%) capacity gains can be achieved using optimum power control. These gains are, of course, not additive with the gains obtained by DCA schemes. However, preliminary results regarding combined DCA/PC schemes show substantial capacity gains[13,15]. For packet communication with short messages or in frequency hopping environments, power control as described above may not work properly due to fact that the feedback delay in the power control loop may in fact be longer than the time required to transmit the message (or in FH the chip/burst duration). The bursty interference caused by other users compounds to the problem of accurately measuring and predicting the C/I. Several approaches to this problem have been proposed. In systems utilizing mainly Forward Error Correction, C/I-balancing power control

11(20)

ZANDER, RADIO RESOURCE MANAGEMENT

strategies involving estimating statistical parameters of the C/I, e.g. the average C/I (measured over many packet/chip durations) as well as fast C/I estimation/tracking schemes have been proposed[34,35,36,37]. In random access systems utilizing retransmissions(ARQ), however the picture is quite different. Two important differences between these systems and the continous (“circuit switched”) systems can be noted. Providing equal transmission quality to all packet transmission (e.g. letting all packets be received with equal powers), turns out to be a disasterous strategy, since it guarantees that no packet can be received properly when several packets happen to be transmitted simultaneously(a “collision”). It has clearly been demonstrated that a received power spread, such as the one caused by near-far effects or Rayleigh/Shadow fading actually improves the capacity (throughput) of such systems [39]. Results show that power control that create an even larger power spread can produce even better results[40,41,42].

4. Managing change - the dynamics of resource allocation
As terminals move about in the service area, the propagation and interference situation may turn such that the terminal cannot be supported by the same access port on any waveform. New terminals may enter the service area requiring services, while others are terminating their communication sessions. As most of the basic resource allocation strategies described above deal mainly with static or quasi-static situations that are encountered on a microscopic,short term, time scale, we have to devise resource management schemes capable of handling these variations. In the first case, where we have signal quality variations due to terminal moving during a communication session some kind of resource reallocation may become necessary. This may be a waveform reallocation, or an “intra-port” handoff, which in principle involves a re-execution of the basic channel allocation scheme, or an inter-port ("inter-cell") handoff . In early cellular systems which are mainly noise-limited systems these handoffs were basically triggered by too low received signal levels. The handoff mechanism has, for these cases often been modeled as a selection (macro-) diversity scheme where the terminals is assigned to the access port with the highest received signal level. This situation where hand-offs occur at more or less well defined "cellborders" has been extensively studied.

12(20)

ZANDER, RADIO RESOURCE MANAGEMENT

Maximizing the instantaneous received signal level may, however, not be neither very practical nor produce the best results. In high density wireless systems, the coverage areas of the access port overlap to a large extent. Low signal levels is rarely a problem since normally several access ports provide sufficient signal levels. In these cases the variations in the interference and not cell boundary crossings is to most probable cause of a handoff. When several access ports may provide sufficient C/I, the system is also able to handle traffic variations by means of load sharing, i.e. by letting less loaded access ports support terminals even though they are providing less C/I than the best (often the closest) port[20]. Combinations of power control and access port selection also show promising results[21,22]. Instead of conventional handoff schemes ("switching diversity"), also continuous combining schemes ("soft handoff") have been studied quite extensively[19]. Keeping track of the mobile terminals in a large (possibly global) wireless system, the mobility management, is a formidable task. Although this is handled mainly on the fixed network end, there are important implications to the resource management. The trade-off between the capacity required for the air signalling to monitor the whereabouts of the terminals (the "locating" procedures) and the capacity required for finding, or paging, a terminal when a communication request comes from the network end, has received quite some attention[23,24] in CDMA schemes. Handling arriving and departing terminals poses a sligthly different problem. Whenever a new terminal arrives (a new request for service or an inter-port handoff) the RRM system has to decide if this particular terminal may be allowed into the system. An algorithm making these descisions is called an admission control algorithm. Since the exact terminal population and gain matrix may not be tracked exactly at all times and due to the complexity of the RRM-algorithms, determining the success of an admission descision may not be possible beforehand without physically executing the admission itself. The admission procedure may fail in two ways I: (“False admission”) A terminal is admitted giving rise to a situation where one or more terminals cannot be supported (not necessarly including the admitted terminal) . (“False rejection”) A terminal is rejected when successful resource allocation actually was possible.

II:

Traditional approaches involving static channel allocation normally use a simple thresholding strategies on the available channels in each cell. Access ports have here been assigned a fixed set of channels that “guaranteed” to provide a certain low outage probability. Such a system is what we

13(20)

ZANDER, RADIO RESOURCE MANAGEMENT

call “blocking limited” Whenever a call arrives, we may check if there are channels available or not. Since we may choose to give priority to on-going session experiencing an inter-port handoff, new arriving calls are admitted only up to the point where there is only some small fraction of the resource remaining. This spare capacity is “reserved” for calls entering a cell due to an inter-cell handoff. In systems using dynamic channel allocation or random allocation there is no clear limit on number of channels/waveform that can be used. In such “interference-limited” systems, the feasibility of admitting new users will depend on the current interference situation. In particular in systems utilizing C/I-balancing power control this is complicated by the fact that already active terminals will react to the admission of a new terminal by adjusting(raising) their transmitter powers. It is therefore quite possible that the admission of yet another user may cause that several of the original users (possibly all!) no longer can be supported at the required C/I-level. Admission control schemes can be grouped into non-interactive schemes and interactive schemes[43]. The noninteractive schemes proposed are mainly using different types of interference or transmitter power thresholds[43], i.e. when the measured interference (or the currently used power) on some channel (cell) is too high, admission is denied. The interactive schemes involve the gradual increase of the power of new terminals until they are finally admitted. Such a procedure to protect the already established procedure connection is referred to as “Soft-and Safe (SAS)” admission[44] or channel probing/active link protection[45].

5. Resource management issues in Future Systems
In most of the designs discussed rely on that the systems carries circuit switched traffic of rather moderate data rates (e.g. speech, low rate circuit switch data). Let us now turn to some of the more important features of the traffic expected in future systems and what impact these will have on system design in general and on radio resource management in particular. 5.1 High bandwidth Future systems are expected to require much higher data rates than current systems. In third generation wide area personal communication systems (cf. UMTS, FPLMTS) data rates in the range 64 kbit/s - 2 Mbit/s are discussed. In local area networks speeds beyond 10 Mbit/s and are common practice. Even radio access at ATM (155 Mbit/s) rates has been discussed[29]. Data rates

14(20)

ZANDER, RADIO RESOURCE MANAGEMENT

in personal communication systems are certainly limited by propagation conditions as multipath etc., but the primary constraining factor is the link budget, i.e. terminal power consumption[32]. As one could phrase it: “A picture does not only say more than a thousand words” - it is usually a thousand times more power consuming to transmit that picture. Since the required transmitter power increases linearly with the bandwidth, high speed radio access will have but a very limited range. The latter clearly has repercussions on the economics of such systems: either we will have to invest heavily in a dense ubiquitous infrastructure, or be limited to cover only certain areas were we will expect to find users width extensive bandwidth requirements. The design and performance of RRM algorithms is not affected very much by the increase in bandwidth per se. In fact, much of the on-air signalling required by many of the adaptive schemes will, relatively speaking, occupy a smaller fraction of the available bandwidth. Increasing the infrastructure density, with more RAPs, will clearly cause an increase in complexity in the RRM algorithms. The focus on distributed schemes will be even more pronounced in these systems. 5.2 Bursty data traffic If the bandwidth as such is not that important to the design and performance of RRM algorithms, the traffic characteristics are. In particular data traffic, but also speech and file transfers, can be seen as discontinuous streams of symbols. There are two main problem areas involved: ? Delay constraints

Circuit switching systems are normally designed to meet absolute delay constraints, whereas the delay for data traffic normally is constrained in the statistical sense (e.g. average delays). The latter type of constraints implies an extras degree of freedom in the resource allocation procedure leading to better resource utilisation. We may trade off blocking for additional delay. This has lead to the design of radio access schemes particularly designed for delay non-sensitive, very “bursty” traffic, so called packet radio system. Since messages are short and delay is to be kept low, in these systems there is no time for the exchange of resource allocation information. Instead random access schemes are utilised in which the terminals compete for the radio resources. This will at times lead to message collisions or conflicts which have to be resolve by special conflict resolution schemes (at a delay penalty).

15(20)

ZANDER, RADIO RESOURCE MANAGEMENT

?

Insufficient information Systems with intermittent data transmission, will also suffer from a different kind of problem. Since there are no continuous transmissions, good link quality estimates cannot be made at will but only when there actually is a transmission in progress. In particular when the traffic is very “bursty, the statistical estimates of the link-quality parameters can degrade considerable since the terminal may move a considerable distance between transmissions. This affects all type of RRM decisions, e.g. channel allocation, power control and hand-off decisions. In these situations channel allocation decisions and power control has to be made on estimated average link qualities rather than on instantaneous values. In these cases, the concept of a “hand-off” looses it meaning in the physical sense and one may instead consider different “connection-less” (or multiport) schemes where any RAP in some area may receive messages from a mobile terminal without the explicit establishment of a logical/physical connection[30,31]. Another possibility considered (in particular in CDMA-type systems) is to “artificially” maintain a physical link even when there are no data to transmit by prescribing a minimum “idle” power level. These tradeoffs are, of course, the more important, the more rapidly the terminals are allowed to move relative to the duration of these “idle periods”.

5.3 Mixed traffic The key problem in “multimedia”-type system is the data rates and delay constraints traffic in small cell environments will exhibit very large peek to average capacity demands. Video users with absolute delay requirements may require considerable portions of the spectrum which they share with email-type message traffic with no such absolute constraints. Dynamic channel allocation (i.e. statistical multiplexing) will provide even larger capacity gains in these situations than in today’s mobile phone scenarios. Conventional single cell traffic multiplexing/averaging will hardly be sufficient to handle the large range of data rates envisaged. Dynamic spatial resource reuse has the potential of broadening the traffic basis for efficient use of specturm resourced is of paramount importance[27]. Another problem caused by the traffic mix, is that also this traffic property makes link-quality assessment more difficult. The conclusion above leads us to the design principle to dynamically share all the available spectrum for all types of traffic. A side-effect of this is, however, that also the interference experienced by different users will exhibit the same wide span in character[38]. In

16(20)

ZANDER, RADIO RESOURCE MANAGEMENT

particular if we would like to estimate the link quality for a high quality circuit switched service, the link will be subject to both quasi-constant as well as intermittent interference (from packet service users). Reliably estimating, for instance, the C/I as a basis for RRM decisions, will be considerable more difficult

6. Conclusions & Discussion
Above we have presented a formulation of the radio resource problem based on the three basic allocation decisions: waveform, access port and transmitter power. As the reader may have realized, these are closely related. Most of the recent work indicates that good results are achieved when these decisions are coordiated. Combinations of power control and DCA[13,15], base station assignment and power control[21,22] as well as power control assisted admission schemes have provided interesting results. Another area were research is just in its preliminary stages is the combination of detailed modulation waveforms/channel coding and there interaction with DCA and power control. Although the current mobile telephony systems are rather easily modelled in the terms described above, it seems clear that also most of the RRM problems expected in the future can be mapped onto the framework presented here. A key problem in bursty and mixed traffic was identified, i.e. the trade-off between maximizing instantaneous resource utilization (transmit only when data is available) and obtaining reliable quality measurements to facilitate the efficient adaptation of the radio resources to the needs of the users. Traditionally we consider the frequency spectrum to be the resource to be shared. Since there, in fact, does not exist any upper limit on the capacity that can be provided (with an dense enough infrastructure), it is important that we widen the resource management perspective. Parameters such as infrastructure density costs and terminal power consumption play important roles. One could easily identify trade-offs such as where the signal processing load should be put in a wireless system - in the terminal where power is scarce or in the fixed infrastructure. The key question here is: Should the access port infrastructure be very dense (and costly) allowing for "dumb", cheap, low power terminals or should terminals be more complex allowing for the rapid deployment of a cheap infrastructure at the expense of battery life and terminal cost ?

17(20)

ZANDER, RADIO RESOURCE MANAGEMENT

7. References
[1] [2] Lee, W.C.Y, "Mobile Communication Fundamentals", Wiley, New York, 1993 Furuya, Y., Akaiwa, Y., "Channel Segregation - a Distributed Apdaptive Channel Allocation scheme for Mobile Communication Systems", Proc DMR-II, Stockholm, 1987. [3] Beck, R. Panzer, H., "Strategies for handover and dynamic channel allocation in micro-cellular mobile radio systems", IEEE Veh Tech Conf VTC89, May 1989. [4] Cox, D.C., Reudink, D.O., Dynamic channel assignment in high capacity mobile communication systems, Bell Syst Tech J. , vol 50, no.6, July-Aug 1971. [5] Everitt, D., Manfield, D.," Performance analysis of cellular communication systems with dynamic channel allocation", IEEE Trans Sel Areas Comm, vol 7, no.8., Oct 1989. [6] [7] Halpern, S.W., "Reuse partitioning in cellular systems", IEEE Veh Tech Conf VTC85, May 1985. Zander, J., Eriksson, H., "Asymptotic bounds on the Performance of a class of dynamic channel Assignment Algorithms", IEEE Trans Sel Areas Comm, vol 11, no.3., Aug 1993. [8] [9] Kanai, T, "Autonomous Reuse partinioning in cellular systems", IEEE Veh Tech Conf VTC92 May 1992. Goodman, D.J., Grandhi, S.A., Vijayan, "Distributed dynamic channel assigment schemes", IEEE Veh Tech Conf VTC93, May 1993. [10] Zander, J, "Performance of Optimum Transmitter Power Control in Radio Systems", IEEE Trans.Veh Tech., vol 41, no.1, Feb 1992 [11] Foscini, G.J., Mijanec, Z., "A simple distributed power control algorithm and its convergence", IEEE Trans.Veh Tech., vol 42, no.4, Nov 1993 [12] Grandhi, S.A., Zander, J., Yates, R. "Constrained power control", Wireless Personal Communications, (Kluwer) vol 2, no3,Aug 1995 [13] Foscini, G.J., Mijanec, Z., "Distributed autonomous wireless channel assignment algorithm with power control", IEEE Trans.Veh Tech., vol 44, no.4, Nov 1995 [14] Frodigh, M, "Bounds on the Performance of DCA-Algorithms in Highway Micro Cellular Radio Systems", IEEE Trans.Veh Tech., vol 43, no.3, Aug 1994 [15] Frodigh, M, "Performance bounds for Power Control Supported DCA-Algorithms in Highway Micro Cellular Radio Systems", IEEE Trans. Veh Tech., vol 44, no.2, May 1995. [16] Tekinay, S, Jabbari, B., "Handover and channel assignment in mobile cellular network,IEEE Comm Mag., vol 29, no. 11, Nov 1991. [17) Austin, M., Stüber, G., "Cochannel Interference modelling for signal-strength based handoff", Electronics Letters, vol 30, pp1914-1915, Nov 1994.

18(20)

ZANDER, RADIO RESOURCE MANAGEMENT

[18]

Zhang, N., Holtzman, J., "Analysis of Handoff Algorithms using both Absolute and Relative Measurements", IEEE 44th Veh Tech Conf VTC94, June 1994.

[19]

Zhang, N., Holtzman, J., "Analysis of a CDMA Soft Handoff Algorithm", Proc PIMRC 95, Toronto, Sept 1995.

[20]

Eklundh, B., "Channel utilization and Blocking Probability in Cellular Mobile Telephone Systems with Directed retry", IEEE Trans Comm, vol 34, no.4, April 1986.

[21]

Yates, R., Huang, C-Y.,"Integrated Power Control and Base Station Assignment", IEEE Trans.Veh Tech., vol 44, no.4, Nov 1995

[22]

Hanly, S.V., "An algorithm for combined cell-site selection and power control to maximize cellular spread spectrum capacity", IEEE Trans Sel Areas Comm, vol 13, no.7., Sep 1995.

[23]

Markoulidakis, J.G., Sykas, E.D., "Model for location updating and Handover Rate Estimation in Mobile Telecommunications, Electronics Letters, vol 29, no 17, Aug 1993

[24]

Morales-Andes, G.,Villen-Altamirano, "An approach to Modelling Subscriber Mobilty in Cellular Radio Networks", Forum Telecom 87, Geneva 1987.

[25]

Salmasi, A., Gilhousen, K.S., "On the System Design Aspects of Code Division Multiple Access(CDMA) and Personal Communication Networks", IEEE 42ndVeh Tech Conf VTC92, May 1992.

(26]

Eriksson, H., Gudmundson, G., Sk?ld, J., Ugland, J.K., Willars, P., "Multiple Access Options for Cellular Based Personal Communications", IEEE 43rdVeh Tech Conf VTC93, May 1993.

(27]

Anderlind, E., "Resource Allocation for Heterogenous Traffic in a Wireless Network", Int Symp on Personal, Indoor and Mobile Radio Comm.,PIMRC 95, Toronto, Sept 1995.

[28]

Gudmundson, G., Sk?ld, J., Ugland, J.K., "A comparison between CDMA and TDMA systems", IEEE 42ndVeh Tech Conf VTC92, May 1992.

[29]

Acampora, A, “Wireless ATM: A perspective on issues and prospects”, IEEE Personal Comm. Mag., Aug 1996.

[30]

Robool, C, “Roobol, C, “Message Delay in 1-D Indoor Packet Radio Systems with Diversity Reception”, IEEE First Symposium on Communications and Vehicular Technology in the Benelux, Delft, The Netherlands, Oct 1993.

[31]

Borgonovo, Fratta, Zorzi & Acampora, “Capture Division Packet Access”, IEEE Comm. Mag., Sept. 1996, Vol. 34, no 9.

[32]

Maguire,G.Q, Ottersten, B.,Tenhunen, H., Zander, J., “Future Wireless Computing & Communication”, Nordisk Radioseminarium, NRS-94, Link?ping, Sweden, Oct 1994.

[33]

Olofsson, H, “Interference Dicersity as means for increasing capacity in GSM”, Proc EPMCC’95, Bologna, Italy, Nov 1995.

19(20)

ZANDER, RADIO RESOURCE MANAGEMENT

[34]

Rosberg, Z, “Fast Power Control in Cellular Networks Based on Short Term Correlation of Rayleigh fading ”, Proc 6th WINLAB Workshop on Third Generation Wireless Information Networks, New Brunswick, NJ, March 1997.

[35]

Rosberg, Z, Zander, J, “Power Control in Wireless Networks with Random Interferers”, Internal Report, Radio Communication lab, Royal Institute of Technology, Dec 1995 (to be published, http://www.s3.kth.se/s3/radio/PUBLICATIONS/documents.html)

[36] [37]

Andersin, M, Rosberg, Z, “Time-variant Power Control”, Proc PIMRC-96, Taipeh, Oct 1996 Mitra, D., Morrisson, J.A., “A distributed power control algortihm for bursty transmissions in cellular CDMA networks”, Proc 5th WINLAB Workshop on Third Generation Wireless Information Networks, New Brunswick, NJ, 1995.

[38]

Mitra, D., Morrisson, J.A., “A novel distributed Power Control Algortihm for classes of service in cellular spread spectrum wireless networks”, Proc 6th WINLAB Workshop on Third Generation Wireless Information Networks, New Brunswick, NJ, March 1997.

[39]

Arnbak, J.A., van Blitterswijk, W., “Capacity of Slotted ALOHA in Rayleigh fading channels”, IEEE J Sel. Areas Commun, SAC-5, No. 2, Feb 1987.

[40] [41]

Metzner, J.J,. “On improving utilization in ALOHA Networks”, IEEE Trans Comm., COM-24, April 1976 Leung, C., Wong, V., “A transmit power control scheme for improving performance in a mobile packet radio system”, IEEE Trans Veh Tech., VT-43, No. 1, Feb 1994.

[42]

Roobol, C, “On the Packet Delay in wireless local area networks with access port diversity and power control”, Proc PIMRC-95, Toronto, Sept 1995.

[43]

Andersin, M, “Power Control and Admission Control in Cellular Radio Systems”, PhD Thesis, Radio Comm Systems Lab, Royal Institute of Technology, June 1996.

[44]

Andersin, M, Rosberg, Z, Zander, J., “Soft Admission in Cellular PCS with constrained power control and noise”, Proc 5th WINLAB Workshop on Third Generation Wireless Information Networks, New Brunswick, NJ, 1995.

[45]

Bambos, N., Chen, S.C., Mitra, D.,”Channel probing for distributed access in wireless communication networks”, Proc GLOBECOM ‘95, Singapore, Nov 1995

[46]

Katzela, I., Naghshneh, M, “Channel Allocation Schemes for Cellular Mobile Telecommunication Systems: A Comprehensive Survey”, IEEE Personal Communications, June 1996, pp 10-31.

20(20)


相关文章:
东南大学研究生课程管理规定.doc
Network 《3G 网络的 QoS》 Radio Resource Management for Wireless Network 《...Course Introduction (including teaching goals and requirements) within 300 ...
微蜂窝中高效地最大化具有QoS需求用户数的资源分配策略.pdf
futurewirelessnetworksduetoitsabilitytoboost appealingtechnique to boost resource...interferencemanagement,to Meanwhile.trafficclassiSboominginworkdesignersprovision....
Future requirements for thermal management_图文.pdf
Future requirements for thermal management_电子/电路...tank leads to thermally induced operating limits n... rfc5951.Network Manage... 24页 免费 Energy...
Resource Management Strategies in Future Composite Radio ....pdf
Resource Management Strategies in Future Composite ...requirements for mechanisms that promote cost-...“Emerging mobile and broadband wireless networks”...
Future Requirements and Concepts for Cabins of Blended Wing ....pdf
Future Requirements and Concepts for Cabins of ...Wireless blue tooth like support of mobile ...simple operation limit the comfort of the ...
GPU Technology Trends and Future Requirements_图文.pdf
GPU Technology Trends and Future Requirements_IT/...Network for ROPs (Raster Operation Processors) to...5. The Vcc range narrowing limits the flexibility...
...and future requirements for Thermal Management Simulation ....pdf
State-of-the-art and future requirements for Thermal Management Simulation (2009)_医药卫生_专业资料。European GT-SUITE Conference 2009 page 1 European GT-...
...Issues in Radio Resource Management of Wireless Systems ....pdf
Dependability Issues in Radio Resource Management of...future wireless systems, dependability requirements, ...(CApacity and network management platform for ...
...Radio Resource Management in Multi Standard Wireless ....pdf
velocity ?QoS requirements ?radio interfaces Figure...In the future, if not already today, it can ... Radio Resource Management for Wireless Networks, ...
Adaptive Radio Resource Management for GSMGPRS Networks.pdf
Zander, “Radio Resource Management in Future Wireless Networks: Requirements and Limintations”, IEEE Communications Magazine, Aug. 1997. [3] I. Katzela,...
Cross-Spectrum Resource Management and Access for ....pdf
Cross-Spectrum Resource Management and Access for ...However, the increasing requirements can hardly be...II. HETEROGENEOUS CELLULAR NETWORKS Future broadband...
RECONFIGURATION MANAGEMENT IN 4G MOBILE NETWORKS REQUIREMENTS....pdf
reconfigurability as a technological enabler of future multi-standard mobile systems and flexible joint radio resource management across different wireless ...
...Transport in Future RANs Under Varying QoS Requirements_....pdf
in Future RANs Under Varying QoS Requirements ...radio resource management network-wide to all ...IEEE Wireless Communications, 4348, February ...
...the IDLocator split concept for Future Wireless Networks (....doc
Future Wireless Networks (FWNs)_信息与通信_工程...the subjective quality requirements from human users...radio resource allocation algorithm for orthogonal ...
...Park Management System based on Wireless Sensor Networks ....doc
Wireless Sensor Networks Vanessa W_工学_高等教育_...in future pervasive computing systems as they can...requirements of a car parking management system ...
A Resource Management and Control Model.pdf
Section II.A, a new resource management and control model suitable to the characteristics of future wireless networks, UCCE, is described in this section....
Resource Management for Future Generation Grids_免....pdf
This paper discusses the requirements for and functionalities of resource management systems for future generation Grids. To this end it is also necessary ...
Resource Allocation in Wireless Networks.pdf
MUs use radio channels to communicate with BSs ...This necessitates prediction of both future clients...for network resource management in wireless LANSs....
...of Highly Secure Broadband Mobile Wireless Networks for ....pdf
The Future of Highly Secure Broadband Mobile Wireless Networks for Public Safety_信息与通信_工程科技_专业资料。The Future of Highly Secure Broadband Mobile ...
...during handoff for mobile multimedia wireless networks_....pdf
multimedia wireless networks_工学_高等教育_教育专区...resource requirements of all future handoff requests...radio telephone system with prioritized handoff ...