A number of specific research topics were identified by the workshop participants as having particular relevance to the important issues related to the design, development and eventual deployment of future wireless mobile communication systems and networks. The research topics described here are not claimed to provide an exhaustive compendium of all important and relevant research areas, nor is it intended to discourage the submission of relevant research proposals in areas not specifically discussed. Rather, the topics listed are intended to provide a sense of those areas and topics which were extensively discussed by workshop participants and for which a strong consensus existed for specifically encouraging research activities. Indeed, it was recognized that there are other important areas which were not discussed in any detail at the workshop, either because of time constraints, or due to the absence of appropriately knowledgeable advocates.
Subject to the preceding qualifications, then, the following provides a list of important research priorities identified and discussed at the workshop. The order of topics presented is completely arbitrary and is not intended to imply any priority ranking. Furthermore, the topics are listed as separate and distinct while a strong consensus of the workshop was that there is a great deal of potential for interdisciplinary or cross-cutting research activities which span many of these topics. More will be said on cross-cutting research activities in a later section.
Prior to the workshop, attendees submitted position papers that described their recommended priorities for research in the field. The position papers were discussed, and this discussion formed the basis for the research priorities identified by the workshop. Wherever possible, we have organized the topics into broad classes, attempting to impose order and structure on the wide variety of research issues, problems, and programs brought up during the workshop. This resulted in nine broad areas of research priorities, which are discussed below in more detail:
Information theory provides the basic theoretical and contextual underpinnings for many communications-related research activities, and research in this area is expected to be highly relevant to wireless mobile communication systems and networks. A great deal of the success of recent developments in the design and engineering of high-speed, energy and bandwidth efficient point-to-point communication links can be attributed to the insights gleaned from information theory. For example, design methodologies based upon capacity and cutoff rate arguments have been extremely useful in optimizing parameters of modulation/coding systems and in tradeoff studies by abstracting the basic attributes of typical point-to-point links in terms of a relatively few relevant information-theoretic quantities. Choosing system parameters to optimize these quantities helps to localize the search for appropriate system designs thereby avoiding an otherwise intractable search of a much higher-dimensional parameter space if specific system designs were to be considered. Similar comments apply to information-theoretic based design methodologies for source coding or data compression. Information theory also establishes fundamental limits on achievable performance of channel modulation/coding and source coding/data compression systems which provide useful performance benchmarks in assessing the relative efficiency of various point-to-point communication systems. To be sure, the use of information theory in this sense as a contextual performance assessment and design tool requires some simplified modeling assumptions which need to be expanded to adequately treat the time-varying fading/multipath and interference-dominated nature of typical wireless links. Furthermore, if information theory is to provide a truly useful integrative contextual tool we need to be able to include other interference mitigation techniques, such as space-time processing and multiple-access techniques, within our link models as well as develop appropriate source models for important multimedia sources, such as images and video, which allow an information-theoretic based rate-distortion treatment. Finally, the basic theory needs to be extended to incorporate delay issues which will be particularly important for real-time multimedia services.
Although it is well-recognized that multiuser information theory can provide an analogous tool in the conceptual design and development of wireless networks that single-user information theory provides for point-to-point links, the appropriate theory in this case is much less developed. Nevertheless, some important results and insights have already been obtained although generally under rather restrictive modeling assumptions. For example, multiuser information theory has shed some light on the optimum use of a single communication channel by multiple streams of a layered, or scalable, source and the importance of shared or common information. This has led to generalized time-sharing concepts which offer the potential of significantly improved performance over the obvious brute-force time-sharing approach. The potential performance advantages on channels more representative of the wireless environment have yet to be evaluated. Likewise, important advances have been made in multiuser detection and estimation theory which appear to offer some potential for wireless networks but more realistic channel modeling assumptions are required to assess this potential.
An important result of information theory is the ability to completely separate source and channel coding issues as embodied in the separation theorem. However, this is an asymptotic result appropriate when no constraints exist on complexity. Under more realistic complexity constraints, there may well be distinct advantages to joint source/channel coding approaches. Furthermore, recent counterexamples to the separation theorem have surfaced; these are generally situations where complete knowledge of channel conditions is unavailable. This situation is particularly relevant to wireless point-to-point channels and, as a result, the implications for joint source/channel coding approaches on such channels is unclear. The implications for wireless networks is even more uncertain.
Based upon this assessment of the potential usefulness of information theory to wireless and mobile communication systems and networks, and the corresponding state of development of relevant approaches, a number of specific research topics were identified which include:
Bandwidth-efficient modulation/coding techniques, such as trellis-coded multilevel modulation schemes, have met with unprecedented success on relatively benign channels, such as wireline telephone subscriber loops and wireless satellite links. Sophisticated multilevel signaling constellations when used in conjunction with appropriate trellis coding structures and corresponding soft-decision decoding algorithms have demonstrated performance closely approaching capacity bounds. Because of the extraordinary premium on energy and bandwidth-efficient performance associated with the wireless mobile environment, research directed toward the development, characterization and implementation of analogous energy and bandwidth-efficient modulation/coding schemes for wireless mobile operation is highly relevant and has the potential for significant payoff. However, the signaling constellations, coding structures and decoding algorithms which have been developed for more benign channels are no longer appropriate for representative wireless links.
For example, existing trellis-coded modulation schemes are generally developed to have large Euclidean distance between distinct transmitted signaling sequences. For typical time-varying fading/multipath channels the Euclidean distance is no longer the appropriate design metric but rather distinct signaling sequences should be well-separated in terms of a non-Euclidean metric that depends explicitly upon channel parameters which may vary from time-to-time and from link-to-link. As a result, new coding structures must be developed which specifically address the distinguishing characteristics of the wireless physical layer. Furthermore, this should be done in such a way as to result in a robust approach relatively insensitive to underlying link parameters. Often robust operation can be achieved through some form of interleaving/deinterleaving of the encoded data stream but generally at the expense of delay. For certain services, such as multimedia delivery, this delay may be unacceptable and alternative approaches for achieving robust performance need to be investigated.
Another distinguishing aspect of the wireless mobile environment that has some influence on appropriate choice of signaling constellations is the overriding requirement for low power or energy efficient performance. Existing bandwidth-efficient trellis-coded modulation schemes generally employ some form of high-level QAM signaling constellation. The premium placed on energy-efficient performance may well require some form of constant-envelope modulation format in order to employ more power-efficient nonlinear output amplifiers in the mobile terminals. In this case QAM signaling constellations are no longer useful and research is required to develop and characterize appropriate bandwidth-efficient constant-envelope modulation/coding schemes for operation in the wireless mobile environment. At the same time clustered modulation constellations for hierarchical transmission need to be studied.
There are also some important research issues with regard to appropriate decoding architectures for bandwidth-efficient modulation/coding schemes operating on representative wireless channels which require some investigation. For example, the workhorse architecture of most existing bandwidth-efficient modulation/coding schemes operating on more benign channels is maximum-likelihood sequence estimation (MLSE), generally implemented using the Viterbi algorithm (VA). For representative wireless links the importance of incorporating channel state information (CSI) in achieving bandwidth-efficient performance is by now generally appreciated and reasonably well understood. However, important issues remain such as how this CSI is to be obtained, how it is to be incorporated into low-complexity decoder implementations, how to quantify and characterize the resulting performance advantages, how to implement schemes for operation in packet-oriented transport systems and how this CSI extracted at the physical layer can be used to advantage in higher-level network layers.
Finally, we should note that appropriate bandwidth-efficient modulation/coding schemes for the wireless mobile environment generally need to operate in an interference dominated environment and provide at best only one component of an overall interference mitigation strategy. Other interference mitigation components might well include multiuser detection strategies, multielement array processing and appropriately designed multiple-access approaches such as CDMA. The modulation/coding scheme should be designed jointly to work in concert with these other mitigation techniques with an overall goal of providing energy and bandwidth-efficient performance under well-defined complexity and delay constraints. In particular, methodologies need to be developed to allow assessment of performance/complexity/delay tradeoffs and the corresponding impact on overall system capacity.
The specific research topics identified in this area include:
Compression technology has developed rapidly in recent years and the appearance of well-defined image, video and audio compression standards has facilitated the introduction of new applications and services. The growing interest in multimedia delivery on fixed wireline networks can be expected to precipitate similar interests for multimedia transport on wireless mobile access networks. The unprecedented bandwidth requirements associated with the delivery of multimedia services, and particularly video, are expected to fuel an accelerated interest in compression techniques despite the existence of international standards and the relative maturity of this technology. Furthermore, the distinguishing features of wireless mobile networks create new issues for compression technology which are either not as significant or entirely absent in the case of fixed wireline backbone networks. Nevertheless, appropriate compression techniques for wireless mobile networks will be expected to interact relatively seamlessly with their counterparts on fixed wireline networks. As a result, focused research on compression techniques, conducted within an appropriate systems context, is expected to be an important topic for investigation with high payoff potential for the evolution and development of future wireless mobile communication systems and networks.
While image and audio compression techniques are expected to be important components of future multimedia delivery systems, the discussion at the workshop focused almost entirely on video compression techniques. There were several reasons for this, not the least of which was the strong conviction that since video was the most demanding application it is important to get this right if there is to be any hope of supporting multimedia services on wireless networks. As a result, the discussion that follows concentrates almost exclusively on video compression although some of the issues apply to image and audio compression as well.
First, because of the premium placed on spectral efficiency for wireless networks, a need exists for yet more efficient video compression approaches than provided by existing standards such as MPEG-x and/or H.26x. Perhaps the potential improvements anticipated for second-generation image and video compression techniques can fulfill this need. A second-generation image coding technique is one that instead of operating on individual pixels or groups of pixels at a time, or the approach used in existing first-generation image coding techniques, operates on image attributes or entities organized at higher levels of abstraction. In this sense it combines many of the features of computer vision approaches with source coding in an attempt to move beyond the apparent performance saturation effects associated with current research directions in first-generation image compression techniques. Similar techniques can be applied to interframe coding schemes to develop second-generation video compression approaches. Considerable research is required to develop, investigate and characterize such techniques and to assess their applicability to wireless mobile networks.
A second distinguishing feature of wireless mobile networks that has some impact on applicable video compression techniques is the relative unreliability of transmission links compared to their fixed wireline counterparts. Most existing video compression techniques have been developed under the assumption of reliable transport or, in some cases, storage. While this is approximately the case for high-quality fixed wireline networks, it is totally inconsistent with the conditions expected on most wireless mobile networks. As a result, the need exists for video compression techniques which are not only efficient but provide relatively robust performance in the face of link errors. There appear to be several ways of doing this, or at least mitigating the effects of channel errors, and research is needed to establish promising approaches. For example, passive error recovery approaches offer the potential for exploiting the inherent redundancy in typical video sequences to recognize gross errors caused by channel error effects and replace them by appropriate spatiotemporal interpolations based upon reliably received information deemed uncorrupted by channel error effects. Unfortunately, these schemes tend to break down in the presence of significantly high and sustained channel error effects and/or high spatiotemporal frequency components rendering the interpolations relatively useless. Perhaps some adaptive hybrid combination of active FEC and purely passive error recovery would be useful here although this needs to be applied judiciously to avoid the bandwidth penalties associated with the FEC overhead. This is a promising research area that requires considerable further study.
Another approach that offers the potential for robust performance in the presence of channel error effects is the use of layered or scalable video coding approaches. Here the encoded video data is represented in terms of a number of layers, each resulting in a distinct data stream, which represent different perceptually-relevant components of the video source material. For example, the different layers may have distinctly different tolerances to channel errors or delay and the corresponding data streams can then be handled differently by the network. These data streams can be prioritized in terms of their perceptual relevance with correct reception of the highest level priority stream providing a base-level reconstructed video quality. In the presence of channel errors, adequate communications resources can be allocated to the high priority layers to provide at least this base-level video quality. In this way a relatively robust video compression scheme can be developed which provides graceful degradation of video quality in the presence of channel errors. We will have more to say about differential protection of the different output streams of a layered or scalable video encoder in the discussion of joint source/channel coding schemes to follow. Existing standards, such as MPEG-x and H.26x, have relatively limited scalability properties. Research is needed to develop efficient scalable video encoders, with perceptually-relevant layered architectures, which are capable of providing robust performance when operating over unreliable channels representative of the wireless mobile environment.
Another issue of importance is the modeling and characterization of the typically variable bit rate (VBR) data streams at the output of video encoders appropriate for wireless mobile networks. This information is important in the subsequent development of admission control and traffic policing strategies as well as congestion/interference control techniques and their effect on reconstructed video quality. Other issues discussed were the need for video coding techniques exhibiting improved delay and delay jitter characteristics as well the requirement for low-power, low complexity, implementations.
The specific research topics identified in this area include:
Typically source coding and channel coding are treated as separate and distinct entities. Justification for this approach is generally based upon the separation theorem of information theory which demonstrates that the optimum system performance theoretically achievable is not compromised by such an approach. Unfortunately, this is an asymptotic result which applies only in the limit of increasing complexity or delay or both. Furthermore, recent counterexamples to the separation theorem have come to light which are generally characterized by situations where only limited or imprecise knowledge of channel conditions are available; a condition of direct relevance to representative wireless mobile links. As a result, practical considerations suggest adopting a joint source/channel coding approach. Here, an overall bit budget is to be allocated between representing the source (source coding) and protection against channel error effects (channel coding) in such a way as to optimize an overall performance criterion. Because of the extremely limited bandwidth and relatively severe error effects on wireless links this is expected to result in substantial performance improvements and was considered a high-priority research topic.
While much recent work on joint source/channel coding has demonstrated considerable success, and has shed some light on the potential performance advantages of this approach, considerable work remains in developing effective and robust schemes which can be useful in the wireless mobile environment. In the first place, most of the work to date in this area has been very specific to the choice of source material, source encoder, channel modulation/coding and channel conditions. For example, schemes have been demonstrated that have been optimized for specific source material, and have exhibited unquestionable performance advantages, for a fixed source coding technique used in combination with a specific modulation/coding scheme and for a specific channel model. It is not clear what the efficacy of these schemes are when applied to a wider class of source material let alone other modulation/coding approaches or channel conditions. Much more remains in developing robust, and perhaps suboptimum, approaches which are less sensitive to source material, source coding scheme, modulation/coding approach and channel conditions. Furthermore, this must be done within the context of a layered communications protocol hierarchy where the information passed from one layer to another is strictly defined.
An area considered particularly ripe for productive research was the use of joint source/channel approaches in conjunction with scalable or layered source coding schemes. More specifically, multimedia source material such as images, video and audio can be decomposed into hierarchical perceptually relevant layers each of which has different QoS requirements of the transport service. For example, high spatial frequency components in a video encoder are generally relatively more tolerant to channel error effects than lower spatial frequency components. Likewise, lower temporal frequency components are generally more delay tolerant than the higher temporal frequency components. This general tendency of scalable source coding schemes to provide layer-specific QoS requirements allows considerable flexibility in the design of a joint source/channel coding approach by allowing the coding to be tailored to the perceptual significance of the different layers while attempting to make most effective use of communications resources. Furthermore, there is some indication that this will lead to more robust designs than existing approaches. In particular, if the high-priority layers represent a base perceptual quality level requiring relatively reliable transport compared to the lower-priority enhancement layers, adequate communication resources can be allocated to insure reliable delivery of the base layer and thus a base-level reconstructed quality. This results in a robust graceful degradation property in the face of wireless link degradations caused by time-varying interference conditions or network congestion.
Unfortunately, most of the existing standards for image, video and audio compression have been developed under the assumption of reliable transport and have little or no provision for a joint source/channel approach. An exception is the possible provision of passive error recovery approaches as part of the standard. Generally these take the form of some type of interpolation technique for filling in data that has either been lost or clearly corrupted by channel error effects. However, these approaches are at best only marginally successful in protecting against the pronounced channel impairments typically associated with wireless links. Here, some form of hybrid approach combining passive error recovery with active joint source/channel coding warrants further investigation. Perhaps the most serious deficiencies of existing standards, for source coding, however, are the severely limited capabilities for scalable or layered delivery despite the obvious systems advantages as noted above. These capabilities, when available at all, have generally been provided retrospectively in a relatively ad hoc fashion and result in considerable performance compromises. Consequently, despite the clear and obvious advantages associated with joint source/channel coding schemes designed on the basis of existing source coding standards there does appear to be considerable room for basic research on the development of new layered source coding schemes which lend themselves more readily to a robust joint source/channel coding implementations for wireless mobile applications.
The specific research priorities identified in this area include:
The use of multielement space-time array processing, or smart antenna, approaches offers the potential for significant improvements in overall system capacity for wireless mobile communication systems and networks characterized as dense cellular operating environments. In the first place, they can obviously be used effectively to mitigate the effects of spatially-distributed interference on the basis of known or estimated spatial/temporal characteristics. Furthermore, under appropriate conditions they can be used to achieve performance advantages by exploiting the inherent diversity, either explicit or implicit, associated with representative fading/multipath wireless mobile channel environments. Finally, they offer the basis for improved multiple-access techniques by appropriate beam steering between base stations and mobile users whose location is either known or estimated.
Despite the obvious system advantages associated with the use of smart antennas, and the considerable research activity in recent years, there remain many open research issues related to their use in wireless mobile communication systems and networks. For example, new fast signal processing approaches are required to obtain rapid and reliable adaptation in the presence of severe delay spread, and the associated large eigenvalue spread, often typical of the wireless multipath environment, particularly for operation in urban areas and enclosed buildings. Likewise, little is known about the effects of severe doppler spread, and the associated time variability, typical of wireless mobile operation and research in this area is required. Furthermore, these adaptation algorithms may well be required to operate in a packet-oriented transport environment servicing a large number of highly-mobile users with diverse and often intermittent service requirements. This places further premium on rapid adaptation.
Other issues include the requirement to develop rapidly adapting algorithms with low-power, low-complexity implementations. The combination of high speed and parallel operation generally results in considerable stress on severely limited power budgets associated with wireless mobile operation. This needs to be avoided at all costs and places a high payoff potential on low-power, low-complexity, implementations. Another important issue is how smart antenna techniques can be integrated with other interference mitigation techniques such as multiuser detection, forward error control (FEC) and power control schemes. For example, could the decoded output be used effectively in a decision-directed data-aided approach to improve the adaptation characteristics? The goal here should, again, be to develop an integrated or joint overall approach that results in robust operation while providing relatively efficient performance all subject to appropriate power/complexity/delay constraints. Finally, we should note the potential for some interaction between the smart antenna, generally residing at the physical layer, and mobility or location management, generally implemented as a higher-layer function. How best to exploit this interaction, and the nature of the information to be passed between layers, deserves further investigation.
The specific research topics identified in this area include:
Appropriate physical and data link protocols need to be developed for wireless mobile networks in conjunction with the embedded MAC sublayer and the higher-level networking and/or transport layers. The design issues here are decidedly different from those for the wired backbone and typical wired subnetworks which have been developed under the assumption of relatively stable and reliable transmission media. Because of the rapidly time-varying nature of the wireless environment, and the associated severe error characteristics of transmission links, considerably more attention must be given to the design of appropriate physical and data link layers. This is further compounded by the anticipated requirement of future wireless mobile communication systems to support integrated services delivery with some form of QoS provisioning. In particular, this implies service-dependent QoS provisioning at least through the data link layer with perhaps further end-to-end control through the network/transport layer.
The physical layer will require some form of robust bandwidth-efficient modulation/coding with the ability to adapt to changing channel conditions, perhaps implemented through some form of channel state estimation and tracking. Furthermore, the physical layer will need to be designed and developed in an integrated fashion by combining the modulation/coding with a possible array of interference mitigation techniques including: smart antennas, multiuser detection, power control, etc. Little is known about joint design approaches such as this and considerable research is required here. The joint design problem is extremely complex, is scenario dependent and requires a search of a high-dimensional parameter space defined by specific system parameter choices to achieve near optimum performance. Perhaps an information-theoretic based design and performance evaluation methodology could be useful here in helping to narrow down the range of appropriate parameter choices. This could be based upon capacity or cutoff rate arguments by incorporating idealized models of the various interference mitigation techniques into overall point-to-point channel models. Issues such as this are largely unexplored although are expected to offer the potential for large payoffs.
Because of the basic unreliability of typical wireless links, the data link layer will not generally be expected to provide a totally reliable packet transport service to the higher network/transport layers. This would require inordinate complexity and/or delay. Some form of end-to-end error control/recovery then will probably be required at the transport layer. The data link layer then should provide a packet transport service with sufficient reliability to allow the transport layer to provide appropriate service-specific QoS requirements. Appropriate architectures for the data link layer will in turn depend not only upon the service being transported but the structure of the underlying wireless network. For example, in a lightly loaded picocellular environment where delays are relatively short, some form of ARQ at the data link layer may be possible for some services, such as graphics and/or data, without exceeding overall delay budgets. On the other hand, this may not be possible for a heavily-loaded macrocellular environment, particularly for real-time services such as voice or video. As a result, some form of FEC would be more appropriate in this case.
An appropriate general data link architecture then may take the form of a hybrid ARQ/FEC approach with the choice depending upon the service as well as the structure of the underlying wireless link. In either case, the combination of physical and data link layer appears as a concatenated coding scheme. The nature of the information passed between layers needs to be defined and appropriate implementations investigated. Whether or not interleaving can be used between layers will depend upon the tolerance to the associated delays imposed. Again, this is service-dependent as well as network-specific. These issues need to be investigated as well as implementation issues. The implementation issues include the development of appropriate fast signal processing algorithms and their efficient low-power realizations in hardware.
The specific research topics identified in this area include:
The development of efficient and reliable wireless mobile communication systems and networks requires a sound understanding and characterization of the underlying RF or IR transmission links and the associated interference environments. This is particularly true at the physical layer but has a ripple effect throughout all layers. It is important then that appropriate channel models and parameter measurements be available in a form that can be effectively used in the development of design approaches.
While considerable work has been done in this area, much of this has been concerned with the measurement of gross channel parameters, such as path loss profiles, shadowing effects and delay/doppler spread parameters under some rather simple and generally stationary point-to-point channel modeling assumptions. Furthermore, even these measurements have generally been obtained for selective wireless bands, such as the wireless cellular band. In particular, there is very little information in any case concerning the distribution and characterization of interference environments. It appears that considerably more work is required in this area which has such an important impact on the design and operation of typical wireless networks.
One area considered potentially ripe for new research approaches is the development of improved channel models that more adequately capture the time-varying behavior of representative wireless links. Another area of considerable importance is to attempt to develop models and associated parameter estimation approaches that can capture the spatiotemporal behavior on wireless links. The latter can be based on a vector-measurement approach utilizing appropriately calibrated multielement antenna arrays. This information would be particularly useful in the development of interference mitigation approaches based upon the use of smart antennas.
The specific research topics identified in this area include:
The provision of spectrally efficient and reliable multiple access techniques for the support of a dense and highly mobile user community operating in the time-varying frequency-selective wireless environment provides a formidable technical challenge. The issues are somewhat different for wireless local area networks providing reasonably high data rates ( > 10Mb/s) over short distances than for cellular systems operating at lower data rates ( ~ 1-2 Mb/s) over longer distances. Introduction of satellite-based personal communication systems (PCS) introduce yet another set of issues. Nevertheless, the common objective is to make the most effective use of the limited spectrum available while supporting a transport service that adequately meets the QoS requirements of the various applications accessing this service. Furthermore, the multiple access technique in use on any wireless subnetwork should allow flexible and efficient internetworking with other wired or wireless subnetworks, however disparate, as well as the wired backbone network.
Multiple-access techniques for existing wireless cellular networks have evolved from the early analog FDMA systems through more sophisticated digital TDMA and CDMA approaches. Each of these latter approaches are appropriate under different circumstances and exhibit distinct performance/complexity tradeoffs, some of which are still not entirely understood. Evolving digital cellular systems have adopted one or the other of these two techniques so, to provide universal coverage for mobile and nomadic users, future mobile terminals will be required to support both approaches in a dual-mode terminal, Nevertheless, the common characteristic feature of these multiple access techniques is that they have been designed, developed and implemented to support a single service - voice messaging - with provision for data services as an overlay feature and little or no capability for QoS guarantees. Future wireless mobile communication systems and networks, on the other hand, will be expected to provide integrated services transport, including multimedia, and will be expected to provide flexible QoS support analogous to that provided on wired backbone networks. These requirements have important implications on the choice of an appropriate multiple access technique and considerable work remains in this area to identify, characterize and resolve the important research issues.
First, in order to improve spectral efficiency, and hence overall system capacity, multiple access techniques need to be closely integrated with various interference mitigation approaches such as use of smart antennas, multiuser detection schemes, power control strategies, channel state tracking and modulation/coding techniques, generally implemented as lower-layer functions. Furthermore, appropriate performance evaluation methodologies need to be developed to allow fair assessment of the resulting spectral efficiency and its impact on overall system capacity.
In addition to its integration with lower-layer functions, an efficient and reliable multiple access approach for future wireless mobile networks should be closely integrated with higher-layer functions as well. For example, the mobility management function, generally implemented at some level above the MAC layer, can provide useful location information to a multiple access scheme based upon the use of smart antennas to mitigate the effects of spatially distributed interference. Likewise, useful information on source state activity can be incorporated into a multiple access approach much like speech activity is used now in CDMA schemes. Perhaps most importantly, an appropriate multiple access technique for future wireless mobile networks with integrated services transport capabilities should be designed to support scalable or layered source coding approaches with layer-specific QoS requirements. Existing TDMA and CDMA schemes cannot provide such features. Furthermore, in order to provide QoS guarantees it is clear that the multiple access technique should be adaptive to changes in interference conditions and user mobility. Again, this can best be accomplished by integrating the multiple access function with information provided by lower and higher levels in the protocol stack.
Finally, since the multiple access function is directly responsible for scheduling access to communication resources it should be closely integrated with the resource management function at whatever layer it is located. This would include system-wide dynamical channel assignment strategies and soft handover capabilities.
The specific research topics identified in this area include:
Although protocols for use in wired networks abound, it is well known that these protocols cannot provide adequate functionality or performance in wireless networks. As wireless networks have much higher error rates, lower bandwidth, and more-frequent outages than wired networks, attempts to use wired-network protocols have resulted in disappointment. For example, the performance of conventional TCP when a wireless link is in the end-to-end path is poor because TCP'’s congestion control and avoidance algorithm works best for networks that experience low packet loss. Mobility also presents a challenge to current protocols, because they were designed primarily for fixed hosts, and routing packets to mobile hosts requires new mechanisms. Because wireless base stations and mobile routing nodes are generally less capable that wired-network counterparts, wireless networks often suffer from excessive delay and latency. New protocols, or extensions to existing protocols that take mobility and the wireless environment into account, must therefore be designed, studied, and prototyped. Moreover, these protocols need to operate in both wired and wireless networks, if they are to be usefully integrated into future networks.
New classes of wireless networks, such as the so-called ad hoc or infrastructureless networks, have no counterpart in today'’s networks. Basic assumptions in current networks, including the notions of a quasipermanent fixed topology and symmetric links, might not apply to such new networks. Accordingly, researchers will have to gain a better understanding of how to design the data link, network, and transport protocols for these networks.
As wirelessness and mobility challenge the assumptions used to design protocols for wired, fixed networks, so too will wirelessness and mobility open new types of applications. While it remains to be seen exactly which novel applications will emerge, one can be assured that applications that incorporate location or position information will require special support from routing and configuration protocols. Likewise, multimedia applications are expected to find special niches in wireless mobile networks.
Given the experiences with Mobile IP and Wireless ATM, it was noted that wireless mobile networks strain the current protocol architectures.
The specific research topics identified in this area include:
Multimedia is expected to be a major traffic source on future wireless mobile networks, and the design of appropriate application-specific transport protocols will play an important role in the provisioning of such services. These transport protocols must interact seamlessly with their peer counterparts on the wired backbone network as well as wired subnetworks. The design of gateways is considerably simplified if the transport protocols used on both the wireless and wired networks are the same. Since ATM is expected to be the transport and switching technique used in the evolving wired B-ISDN backbone, there is a clear advantage to the use of ATM on the wireless network as well. Furthermore, the availability of ATM switches and multiplexers to interconnect base stations simplifies the problem of interfacing to the wired backbone. Despite these apparent advantages, there exist strong arguments in favor of the development of wireless-specific non-ATM based transport protocols. The main argument relates to the relatively large overheads associated with ATM cell sizes and the associated spectral inefficiency when used on severely bandwidth-limited wireless links. Other issues concern the complexity of synchronizing burst-mode radio modems in the fading-dispersive wireless environment. These issues are not completely understood and the relative advantages/disadvantages of ATM-based wireless transport compared to alternative approaches requires much further study.
Regardless of the wireless transport protocol in use, the transport layer must utilize the services provided by the data link layer to provide end-to-end service-dependent QoS guarantees. In addition, congestion and flow control will generally be provided at this layer working in conjunction with the MAC sublayer utilizing information passed through the data link layer. The design issues associated with the transport layer become particularly important for multimedia delivery; e.g., voice, images and video. The multimedia data will typically be provided to the transport layer in compressed form, perhaps using a scalable or layered compression approach, in the form of multiple data streams each with their own QoS requirements. Furthermore, there will generally be strict requirements on the relative latency of these different streams with respect to each other. The transport layer must then provide reliable end-to-end transport of these multimedia data streams, within appropriately defined QoS guarantees, while meeting strict relative latency requirements despite the relatively unreliable packet or cell transport provided by the data link layer. More specifically, the transport layer must operate reliably in the face of errored or dropped cells on the wireless subnetwork, due to noise, fading and interference effects, as well as the tandem effects of dropped cells on the wired backbone network due to congestion. This places considerable burden on the transport layer and little is known at present concerning the design of efficient, flexible and robust protocols for accomplishing this. This is a research area that deserves considerable attention.
In order to provide reliable operation then, an appropriate multimedia transport protocol must provide, in addition to segmentation and reassembly of compressed data streams to/from cells, some form of FEC across cells perhaps combined with some type of cell interleaving. Generally this is performed in an adaptation sublayer and must be carefully designed to avoid introducing additional latency in the form of cell delay and jitter, either of which could be catastrophic to a real-time multimedia service such as video. Furthermore, the design of the adaptation sublayer is further compounded by the fact that the different layers or streams of a multimedia source have different QoS requirements and thus different adaptation sublayers may be required for each source layer. It is not clear how to accomplish this subject to reasonable complexity limitations. This is another research area that requires much additional attention.
Finally, it should be noted that regardless of the wireless transport protocol used, it will be some time before an appropriate repertoire of native-mode applications are available and able to internetwork with existing wired infrastructure. As a result, one can expect that existing TCP/IP applications will be expected to be supported over future wireless networks for some time into the foreseeable future and will provide the major internetworking approach. It is important then that careful consideration be given in the development of future wireless transport protocols to the ability to efficiently support TCP/IP applications. The issues here are not well understood and require careful investigation.
The specific research topics identified in this area include:
To cope with the highly dynamic behavior associated with the wireless environment and mobility, it is widely recognized that protocols should be able to adapt to a wide variety of situations. While protocols in the wired network also adapt to underlying conditions - usually at connection-setup time but sometimes during a connection - the range of adaptation is typically narrow. Response to congestion is common, but the algorithms employed are limited by the need to preserve simplicity. It is fair to say that once a connection in a wired network is established, the underlying conditions will remain relatively stable, save for occasional congestion. This is often not so in wireless mobile networks. The wireless link experiences a range of conditions not encountered on the wired link, e.g., fading, transient service outage, high error rates, burst error patterns, and highly unpredictable traffic on shared links. Furthermore, mobility exacerbates the situation by introducing handovers, motion-induced effects, rerouting actions, and limited battery life.
Adaptive protocols provide fertile ground for advanced protocol research. As nearly all protocol research has been done on relatively static protocol architectures, there is much to learn about how to select a different protocol on the fly when the original one no longer provides the required level of service. Topics of study include how to decide when a new protocol is needed, how to identify the right protocol to be used under prevailing conditions, and how to coordinate a handover. It is also necessary to understand the problems one might encounter with adaptive protocols, such as excessive overhead or thrashing from switching indiscriminately. Nor are the semantics of protocol adaptation well understood. If state is associated with a protocol, should an adaptation require a simple translation of the state, or can an entirely new state description be employed?
Protocol adaptation may be realized in several ways. Active networking, in which packets may contain executable instructions (in addition to headers and data), provides one approach to implementation. The adaptations offered by such an approach are limited only by the expressiveness of the executable instructions to be interpreted by the receiving node. The efficient implementation of adaptive protocols in both hardware and software present interesting research problems. Seeing as how many host computers now offer multiple network interfaces, one can ask how an end-to-end protocol might be implemented that dynamically switches from one interface to the other, transparently to the application and its user. Similarly, an appropriate protocol stack must be specified and constructed, so that adaptivity is supported easily.
The specific research topics identified in this area include:
Wireless mobile networks and communication systems have been subject to very little measurement and experimentation up to now. The progress of research and development in the field demands firm quantitative data, which can be obtained only by thorough empirical study of real networks. Intuitive and anecdotal characterizations of the wireless environment and mobility must be supplemented by more-accurate characterizations if rational design decisions are to be made by researchers. Carefully executed measurements will provide the foundation for sound work in the future. The workshop participants expressed a strong desire to see experimental work in wireless mobile networks encouraged.
Being a relatively new aspect of networking, mobility is not well understood. The movement patterns of users drive the models of mobility upon which important designs, algorithms, and protocols are predicated. Random, analytically generated movement patterns are unlikely to capture the nuances of real mobile hosts accessing an actual mix of applications. The arrival and departure rates of users to and from cells as well as the duration of their sojourns are important statistics that would prove valuable to researchers. Information about movement correlations would also be useful. Such statistics would be relevant to both individuals and groups. As few users currently use mobile hosts, mobility statistics are expected to evolve over time. Indeed, new applications will attract additional users who will employ wireless mobile services in new and different ways. Ad hoc wireless networks, too, will have mobile hosts whose movement patterns will certainly differ from cellular hosts.
As one of the most-critical metrics in wireless systems design, delay and latency occupy a central role. Early measurements of delays in commercial wireless networks indicate that very high latencies can be experienced by the user. Given the sensitivity of multimedia applications to latency and delay jitter, it is important to document and analyze delay performance in wireless mobile networks. The identification of system bottlenecks would permit work to proceed on improving overall performance. Furthermore, the characterization of the components of delay in a wireless mobile system could be used to perform simulation studies.
Traffic measurements also provide needed data for modeling, simulation, design, and implementation. Understanding the extent of burstiness, time-scale dependence, and self-similarity in traffic would help to size system buffers. In the same vein, measurements of error and loss statistics provide the basis for designing error-protection and reliability schemes.
The specific research topics identified in this area include:
Simulation offers a cost-effective way to assess network performance. One of the great challenges of wireless mobile networking is how to build systems that scale. Current techniques do not adequately capture the behavior of extremely large networks. General-purpose simulation systems are needed that allow one to model and design networks and protocols at arbitrary levels of detail. Also needed is the ability to model various wireless environments as well as the mobility of users. Driven by real data, high-fidelity simulations made possible through the use of a rich library of protocol modules can be applied to the thorough evaluation of new network architectures.
Simulation provides the capability to build high-fidelity models of wireless mobile networks. Given the recurring need to improve wireless performance significantly, the development of simulation techniques for integrated modeling of a wireless system (not just a link or a protocol) would have a high payoff. Furthermore, simulation is one of the few means available to researchers to evaluate the scalability of a design before its deployment. Very large-scale simulations that address how a solution scales in various dimensions are necessary to insure that large investments in the wireless infrastructure will produce the degrees of scalability needed to support burgeoning populations of users.
Tools for planning wireless installations are indispensable to network designers and operators. Achieving the coverage and quality expected by users requires the solution of diverse problems. Tools should be developed for use in pico-, micro-, and macrocell systems.
The specific research topics identified in this area include:
Layered between the application and the protocol interfaces, middleware for wireless mobile networks will grow in importance as systems proliferate. Middleware comprises software that can support network transparency and well-defined semantics for existing and new applications. As such, it is a powerful tool for dealing with heterogeneity. Functions provided by mobile middleware include anticipatory prefetching of information, caching of frequently requested data, filtering parts of a data flow in response to QoS fluctuations, and error concealment. Agent software and active code can be used to implement middleware, providing flexible, dispatchable services for remote execution on network platforms. Proxy servers are also handily realized this way. Applications can thus rely on middleware to be a bridge to the primitive services presented by the wireless mobile network.
Middleware derives much of its utility from the abstractions it presents to the application programmer. The abstractions to be used by mobile-application programmers will doubtlessly differ in important respects from the abstractions deemed necessary for normal network-based applications. Questions surrounding this issue include not only which mobile-middleware abstractions are appropriate for the user but also whether mobile middleware can be effectively integrated with ordinary middleware. The application programmer interface (API) to mobile middleware therefore needs to be carefully constructed in accordance with the requirements of mobile applications and with an eye towards using as a base the common middleware packages that are or are expected to be dominant in the marketplace.
Network middleware in general, and mobile middleware in particular, have not been extensively studied, and we have scant experience with the design of mobile middleware. The question of how to cache information for mobile users in the most-effective manner is not answered, and much work remains in this area. The issues surrounding the placement, selection, and access of replicated or mirrored information repositories that can serve a mobile user traveling at high speeds and over great distances deserve deeper investigation and analysis. Replacement policies for cached data that is used by mobile users will differ from the policies used in today's’ wired networks; it would be desirable to develop middleware facilities that can cope with the caching of data for both fixed and mobile users.
Along with added functionality, middleware potentially adds delay. Given the strong need to reduce delay in wireless mobile networks, one must take care to design and implement mobile middleware so that performance is not sacrificed. Techniques for designing high-performance mobile middleware must be developed. Equally important is the need to conduct experimentation with mobile middleware, in order to characterize and optimize its performance. Design and analysis techniques used for gigabit CORBA (Common Object Request Broker Architecture) could be applied to mobile middleware; on the other hand, entirely new architectures for mobile middleware might be called for to account for the differences between the wired and wireless environments. Mobile middleware offers the capability to disguise or conceal errors that are encountered on wireless links. Middleware can condition or adapt to wireless links in ways that link- and physical-layer entities cannot.
The use of middleware to manage wireless mobile networks is also of interest. The administrator of a wireless mobile network often does not enjoy physical access to mobile nodes, base stations, and other devices, and a roaming host might not even be considered part of the domain it is visiting. Remote management of inaccessible resources or visiting hosts is facilitated by mobile middleware.
Whereas the use of intelligent software in communication protocol layers has not been extensive, there is reason to believe that a greater degree of intelligence can and should be incorporated into middleware layers. Decision-making to support smart caching, mobility awareness, error concealment, packet-by-packet protocol selection, resource discovery, and other link-enhancing adaptations fits naturally into mobile middleware layers.
The specific research topics identified in this area include:
Wireless mobile networks are expected to spawn specific applications unique to these networks. One can only imagine at this time some of the applications that might take hold. Nevertheless, researchers can play a special role by creating and prototyping leading-edge applications that are based upon mobility. Besides presenting new opportunities, mobility also burdens applications by invalidating common assumptions of uninterrupted connectivity and continuous update of data. The mobile user will frequently disconnect from and reattach to a network (because of movement, power conditions, or loss of service). It is desirable that the mobile user'’s disconnection and reattachment will be essentially transparent to the user.
There are limitless possibilities for mobile applications. While potential applications exist in commerce, education, medicine, government, public safety, and numerous other areas, market and social forces will determine which are accepted or rejected. Nevertheless, the active contemplation of new mobile applications is an effective tool for researchers to use in guiding their work. Systems and protocols are designed and implemented to support specific applications, and this principle should be actively employed in the wireless mobile networking arena.
The specific research topics identified in this area include:
Location-management functions make it possible to access the network independent of the user'’s location. Not limited only to users, it is easily imagined that entire networks might one day be mobile as well, e.g., networks on aircraft or other vehicles. It should furthermore be possible for users to retain their assigned addresses, so that a connection continues to work even after temporary disconnection and reconnection, and access to network-based resources and services should appear similar regardless of where the user accesses the resource from. Multiparty connections between any combination of mobile or fixed users should also be possible. Location management works at several layers and is, therefore, a complex process.
The emergence of Mobile IP as the dominant protocol for supporting mobile data networking is a productive springboard from which to explore and develop wireless mobile networking. Wireless ATM is also being developed, and its approach should also be the subject of research and investigation. The objective is to understand the fundamental assumptions underlying mobility and the algorithms and protocols that can be used to manage mobility most efficiently. The basic issues of hard and soft handover, registration, paging, QoS routing, multicast, and security are important. It is critical that researchers develop analyses and experiments that support the development of an overall architecture for location management in wireless mobile networks.
The need for the mobile user to register with an agent raises the question of how best to effect the registration. Registration in a cellular system that supports packet switching can occur at both the link and network layers, so it is reasonable to inquire whether it is possible to combine these different types of registration. Handover also involves the link and network layers, and it might be possible to use information from the link layer to effect rapid handovers at the network layer. Given the large delays associated with handover in Mobile IP, new approaches merit consideration. New wireless networks based on microcells will have users who change cells frequently, and efficient methods to register and handover these users will have to be developed.
The tunnel-based triangle routing of Mobile IP is susceptible to breaches of security. Unless measures are taken to authenticate a mobile host, it may easily masquerade as another host, gaining broad access to the home network of the impersonated host. A host can also hijack the session of a legitimately authenticated mobile host, as the foreign agent welcomes a wide variety of mobile hosts. Besides keeping out unwanted intruders, firewalls, the most commonly used defense in enterprise networks, can stop mobile hosts from accessing their home networks. The practice of filtering on source addresses can be incompatible with the tunneling approaches used by schemes such as Mobile IP.
Location management in ad hoc wireless networks is even more difficult. The problem of how to register a mobile node with the appropriate wire-network-attached agent must be solved. The analog of handover in the cellular system is how to determine when and how to register with another agent after the node (or other neighbor nodes) has moved. Also a problem is how to support QoS and mobility in ad hoc wireless networks.
The specific research topics identified in this area include:
Along with the challenges of mobility come new opportunities. Possessing a unique position, a mobile node can seek or offer resources on the basis of where it is located. A mobile node could request a resource or service (e.g., a printer) that is closest to its current or expected future location. Likewise, applications could use the mobile node'’s position or travel itinerary to access cached data or to search for information about specific geographical areas; queries about a service within a given radius of a specified position could be directed to a limited number of servers in the region without flooding the entire network.
Elevating position to the rank of a first-class network object enables many new kinds of service, from transparent pointcasting (newspaper delivery to one'’s “doorstep, wherever it might be) to emergency services (distress calls that identify the caller'’s position). By revealing a node'’s expected location at a future time, it would be possible to offer prestaged delivery of specific information without requiring the node to request explicitly the information upon arrival. Such prefetching can greatly reduce the user;’s perception of latency.
Knowing the position of nodes within a network makes it possible to route geographically by determining a path to the destination as one would navigate by dead reckoning. The proliferation of global positioning system (GPS) receivers makes it possible to obtain accurate spatiotemporal fixes of stationary and mobile units, opening up the possibility of tracking and locating users and resources in the network and establishing landmarks or beacons within the network. At a time when the sizes of routing tables are growing rapidly to accommodate the constant addition of new address prefixes, such an approach offers another alternative to scaling the network. Large address spaces permit us to assign a unique address to every lightswitch in the world but represent a configuration and network-management nightmare. Geographic addressing could again prove useful in this context, especially where addressed entities are mobile.
The right to access a resource can often be linked to one's geographic position. Admission to an office or residence frequently bestows upon the admitted party the right to use specific resources (phones, copiers, etc.). This concept extends easily to mobile networks as well. Equally applicable might be the use of position to enforce export restrictions.
The specific research topics identified in this area include:
Wired and wireless networks both must provide the QoS support required by applications. Although current wired networks provide limited QoS, it is universally agreed that a richer set of QoS levels is needed to allow emerging applications to use these networks. In particular, multimedia applications, which exchange data, voice, images, and video, make exacting demands on the QoS delivered by a network. Although significant research and development efforts are now targeted at how best to provide QoS, there is no consensus on the right approach. Research in providing QoS in wireless mobile networks is necessary to ensure that advanced applications will be supported and that wireless and wired network QoS principles will be compatible and interoperable.
QoS in a wireless mobile network is substantially more complex than in a purely wired network, because the wired network deals with integrated services in a relatively static manner. It is sufficient to establish or guarantee that the end-to-end path has adequate resources to deliver an information flow between end users with the negotiated QoS. With resource reservation one can prearrange for resources to be dedicated for the duration of the flow, thereby guaranteeing the bandwidth, delay, and error rates contracted for by the application. Such a contract remains in effect so long as no resource fails. The situation in a wireless mobile network is much more volatile, because a mobile node'’s end-to-end path is likely to change as it moves through the network, and the wireless link is subject to wide fluctuations in performance and reliability. Resource reservation is thus complicated by the need to consider not only resource availability on the initial end-to-end path but also on other paths likely to be be used as the node'’s position changes. Research in techniques for reserving resources in such a way as to minimize the probability that a flow will be blocked when it is rerouted in response to a node'’s movement is essential in cellular and ad hoc wireless networks.
Adaptation as a means to maintain a specified QoS when the wireless link fluctuates or degrades is an important subject of research. A fundamental understanding of how to trade off processing for bandwidth is a key objective in this work. New algorithms and techniques must be evaluated and compared under realistic conditions.
Different approaches to QoS in wired networks, coupled with possibly new ways of providing QoS in wireless mobile networks, make necessary the consideration of how to map QoS parameters from one framework to the next. For instance, supporting QoS in a network that uses both ATM and IP requires mapping between the different notions of CBR, VBR, ABR, and UBR services in ATM and best-effort, controlled-load, and guaranteed services in IP. It is possible that new service abstractions and mappings will be appropriate for wireless mobile networks.
The specific research topics identified in this area include:
Already important in wired networks, multicast is expected to assume an important role in wireless mobile networks, especially in support of multimedia transmission. Additionally, several fundamental services in wireless mobile networks will be built upon multicast. In particular, multicast is essential for many collaborative systems, in which information is usually disseminated from a source to several destinations. When connected to the wired network, the wireless mobile network adds several twists to multicasting by allowing for node mobility and by using low-bandwidth, unreliable wireless links. Multicasting will probably play a role in supporting features such as caching and prefetching. Position-based multicast, which differs from traditional multicast'’s explicit use of group membership to determine who receives a packet, would provide for the ability to address a packet to all nodes in a specified geographic region. In all cases, management of multicast group membership is nontrivial.
Multicast routing in ad hoc wireless networks poses unique problems not encountered in much of the existing work on multicast routing. This class of network should be carefully analyzed with respect to multicast routing, and appropriate solutions need to be proposed and evaluated.
Multicast routing in wireless mobile networks should able to handle multicast groups of varying densities. The expected sparseness of users in the wireless portion of the overall network makes it necessary to consider techniques that rely on widespread flooding to build routing trees.
How to support reliable multicasting in the wireless mobile network deserves serious attention. When one considers the difficulties associated with reliable unicasting in the wireless mobile network, it is clear that this area requires thorough investigation. The phenomenon of “NACK implosion” threatens to degrade performance unacceptably in the wireless mobile network, and approaches for avoiding this phenomenon in the wired network might be susceptible to poor performance in the wireless mobile network. The behavior of reliable multicast protocols in the presence of mobility and link unreliability has not been studied extensively, and such an undertaking is essential to designing reliable mobile multicast protocols. Moreover, different types of traffic expect different types of reliability, e.g., user-viewed video streams might tolerate relaxed reliable delivery in lieu of reduced latency.
The support of QoS in mobile multicast is challenging when one considers flows that traverse both the more-reliable wired network and the less-reliable wireless network. Must the techniques being developed for the wired network be augmented or modified to work well in the wireless mobile network? The use of reservations, class-based queueing, weighted fair queueing, and other techniques should be investigated in this context.
Also necessary would be the ability to support anycasting as well as multicasting, especially since this is a planned feature of IPv6.
The specific research topics identified in this area include:
Wireless networks are typically viewed as bandwidth-constrained relative to hardwired networks. However, for the portion of a wireless network consisting of battery-powered mobile nodes (this may be only the subset of a wireless LAN or PCS system, or much - if not all - of the network in an all-wireless, multihop, infrastructureless network), a finite energy capacity may be the most significant performance constraint, and thus its utilization should be viewed as a primary network control parameter.
Minimizing energy usage impacts protocol design at all levels of network control. Most naturally, one thinks of minimizing the "on" time of a given transceiver, much in the way present day paging systems maximize the time between battery changes by imposing a duty cycle on the pager. This includes both transmission and reception, as it takes energy to simply monitor a channel. Modern pagers are typically in "receive" mode only a fraction of the time, "sleeping" the rest of the time according to a predetermined schedule. Techniques such as these are relatively easy to employ in systems where the system coverage area is partitioned, each area "centrally" controlled by a given base station as in paging or cellular systems. However, in systems relying on asynchronous, distributed control algorithms at all network levels (including multiple access) such as in infrastructureless networks, participation in the control algorithms prohibits usage of simple predetermined schedules, and more sophisticated methods are required.
It is clear that many wireless network architectures are inherently "asymmetric" from an energy perspective. Base stations may typically be assumed to be to be power-rich, whereas the mobiles they serve may be power-constrained. Thus, this asymmetry should be accounted for in the network protocol design at all levels, offloading complexity from the mobiles to the base stations as much as possible. Again, the problem may be more difficult in infrastructureless networks as the entire network may be energy-constrained. Here, complexity must be uniformly distributed throughout the network (at all levels of protocol operation), and kept as low as possible while simultaneously maximizing the throughput/energy efficiency (measured in bits/second/joule) subject to an upper bound on latency. Note that a standard metric of wireless system design, viz. spectral efficiency (measured in bits/second/hertz), may not be the most significant constraint in power-constrained systems. From an energy-usage perspective, it may be better to be less spectrally efficient.
The core issue which surfaces in energy-constrained systems is that more efficient power usage translates directly to better performance. Multidimensional tradeoffs between energy usage and various performance criteria exist. One may choose to burn power to decrease latency, increase throughput, achieve a higher level of QoS support or some combination thereof.
The specific research topics recommended in this area include:
As an alternative to cellular networks, infrastructureless or ad hoc networks are designed for use in areas where there might not be permanent signal coverage. These are networks in which the transmission range of a node is limited and the topology of the network is dynamic, so that multihop communication is necessary for nodes to communicate with each other. The reliance on wireless multihop communication to maintain connectivity among nodes places added complexity on the design and operation of these networks. The dynamic nature of the infrastructureless network challenges current routing techniques. The problem of how such a network initially organizes and responds to node mobility and terrain-dependent signal degradation requires solution. It is forseeable that nodes will have different transceivers and antennas, and these differences must also be considered when setting up an infrastructureless network.
Proponents of ubiquitous computing contend that computing devices will be embedded in our living and working spaces. Rooms, corridors, vehicles, and public areas might one day each contain hundreds or thousands of computers, which can be interconnected most effectively by means of a wireless network. Although most of the embedded devices would be fixed, it is expected that devices embedded in movable furniture, equipment, or vehicles will be considered mobile or nomadic. Furthermore, users will certainly move from area to area, expecting to have instantaneous connectivity to local embedded devices.
While ubiquitous computing seeks to provide wireless access to information for the peripatetic user who wanders from room to room, micronetworks embed small devices, such as microelectomechanical systems, in hostile or inaccessible environments. As the nature of the environment often makes a wired interconnection impractical or too costly, a wireless network is typically the preferred regime for interconnecting the embedded devices. Such devices might be manufactured into building materials that are used to construct an edifice or vehicle, endowed with special-purpose functionality, such as the ability to monitor strain gauges or accelerometers and pass the information on to a central controller. Given that the embedding matrix itself (e.g., the frame of an aircraft) might be mobile, the micronetwork is also mobile from the point of view of a global backbone wired network. Protocols for initialization, reconfiguration, and self-organization of micronetworks remain to be designed, analyzed, and validated. Under the severe power constraints found in micronetworks, lightweight power-smart protocols are an important topic for research.
Management and control of infrastructureless networks is an important area that will probably draw upon different solutions than are found in either cellular or wired networks. The need to report on a dynamically changing topology suggests a more-autonomous style of network management than one finds in today'’s polling-based systems. Also, the sheer number of nodes that one might find in a ubiquitous computing network or a micronetwork underscores the need for a level of scalability not commonly present in most approaches to network management.
The specific research topics identified in this area include:
Usually limited to end-user nodes, programmability is being pushed into the core of the network. Programmable switches and base stations enhance the flexibility and intelligence of the network by allowing a richer collection of algorithms to reside inside the network. Programmability pertains to the ability of users or third parties to program core nodes over a relatively long period of time (i.e., customization of nodes) or over an extremely short period of time (i.e., active messages executing in nodes). This departure from the traditional concept of persistent protocols is being actively explored by researchers in wired and wireless networking. Specifically, the problems of network programmability in wireless mobile networks is a topic for further research.
Central to network programmability is the definition of an appropriate application programmer interface (API) for the node. A powerful API gives the user the ability to create new services on a node, which is an essential feature for adaptive wireless protocols. The programming environment, whether for persistent or active code, has unique requirements for protection, security, and safety. Moreover, the difficulty of constructing performance-oriented programs demands that support for efficient code be provided.
The specific research topics identified in this area include:
Wireless networks are used principally as a means of access to resources that reside primarily in the wired network. The wireless access network has evolved over time, including analog and digital cellular telephone networks, wireless local area networks (WLANs), mobile data services, personal communications network (PCN), global system of mobile telecommunications (GSM), future public land mobile telecommunications system (FPLMTS), cellular digital packet data (CDPD), metropolitan wireless services, cordless-telephone systems, satellite systems, and millimeter-wave systems. New systems are expected to emerge over time. The study of new wireless access architectures is a rich area to which researchers can make important contributions.
A common goal is to provide ubiquitous access to an integrated data and telecommunications network by means of wireless links. The achievement of this goal implies advancements in the structure of the physical access network and the protocols that provide services on this access network. Simulcast networks, in which messages are broadcast simultaneously across a selected subset of the entire network, are a candidate for use in the access network. Also of interest is a better understanding of how public broadcast systems could be combined with digital data networks. Should all data, broadcast, and telecommunications access points in residences, businesses, and public areas be combined or is it more effective to keep the terminals separate (despite whether the streams are integrated within the network)? The use of hierarchy in the wireless network should also be examined. Ubiquitous access might also merge cellular and ad hoc networks in ways that have not today been foreseen.
The algorithms and protocols to support efficient wireless access need to be designed, analyzed, and validated. These access networks should provide efficient sharing of bandwidth and support for QoS and multicast.
The specific research topics identified in this area include:
There are many aspects to the provisioning of security that need to be addressed as part of the development and deployment of future wireless mobile networks. In the first place there is the rather obvious issue of the confidentiality of communications. A less obvious issue, however, is that of the confidentiality of location information since this information will generally be available through the position location and tracking capabilities associated with the mobility management function. Consideration needs to be given to appropriate means for securing this information.
In addition to the confidentiality issue, there is the important issue of validation of users' access rights in order to deny fraudulent access to unsubscribed or unauthorized network resources. In many ways this is analogous to similar problems on existing fixed wired networks except in one important respect: on many wireless mobile networks access rights to network resources may be highly location-dependent. The validation aspect of network security then must be tightly coupled with the mobility management function. Little work appears to have been done in identifying the important research issues here and this would seem to be an important area for further research.
Finally, an important aspect of network security is the user authentication issue. More specifically, provisions must be made for authenticating the identity of a source of communications. Again, this is an issue that exists, although to a lesser degree, on existing wired networks. The situation is exacerbated on wireless mobile networks due to the relative ease with which fraudulent communications, or spoofing, can be originated. Appropriate network security approaches need to be provided to protect against this form of intrusion. This is an area requiring much more extensive research than is presently the case.
The confidentiality aspect of network security can generally be provided through appropriate encryption techniques. However, many existing encryption techniques are based upon the assumption of reliable network transport. In the presence of transport errors, at least in some modes of operation, they exhibit error extension whereby even a single transport error can result in an extended sequence of errors in the decrypted output stream. This can cause severe problems with certain service classes, such as multimedia data. On error-prone links, such as might be expected in the wireless mobile environment, this problem can be quite severe. As a result, new approaches to the confidentiality aspect of network security are required that exploit a combined communications/networking perspective. This is expected to be a fruitful research area.
The airlink in wireless networks makes security a pressing concern. So too does mobility, which often has the effect of making available to a “foreign” user an enterprise'’s resources. The need for strong authentication of users and resources is clear, especially when a roaming user seeks to access computing resources at the home site. The home site is often protected from the external network by a firewall that discriminates on the basis of addresses, services, and other features of the connection. A mobile user connected to a wireless network or a visited site can communicate freely once it has been authenticated and identified as a “friendly” member of the home site. It is thus obvious that mobility demands a level of authentication not common in today'’s wired networks, which often rely on relatively secure modem connections through the telephone network. Also problematic for security is the airlink exposure, since it is much easier to intercept a wireless link than a wired link. (Even though shared broadcast media such as Ethernet are also relatively simple to intercept, these media typically run through spaces controlled by the owning or operating enterprise. Therefore, it is less likely that corresidents of a common enterprise will snoop on each other, and it is more likely that snoopers will be detected.) Hence, encryption of data to guarantee privacy and confidentiality is essential.
Limited buffering capabilities of mobile nodes and relatively high error rates inherent to most wireless media both result in reordered and lossy packet traffic. Encryption mechanisms must be developed that cope with both reordered and lost packets while providing high security guarantees.
The specific research topics identified in this area include:
Mobility gives users the opportunity to work from different locations. However, it is generally thought desirable that the specific location from which a mobile user accesses the network should remain unknown unless the user wishes to disclose this information. Reasons for this policy are numerous, but the chief consideration is that divulging location information can compromise a user in various ways. The need to protect business transactions from competitors' scrutiny, the rights of people to freely associate with others, concerns about personal safety, and other issues reinforce the importance of this topic. Yet the requirements for anonymity might not always be compatible with other requirements, such as incorporating location-dependent queries into mobile applications. Despite these difficulties, acceptable solutions must be discussed and analyzed.
Related to, but distinct from, anonymity is the requirement for untraceability a user's movements. Fears of surveillance highlight the objective of preventing someone from learning the (future or past) itinerary of another mobile user.
Not strictly technical problems, both anonymity and untraceability impinge upon societal and legal issues. The protection of legitimate rights of citizens must always be weighed against the obligations of society to protect itself against unlawful or harmful actions of miscreants.
The specific research topics identified in this area include:
Each year large sums of money are lost by operators of cellular telephone systems because of fraudulent cloning of users' calling information. Although stronger authentication procedures would help to prevent such fraud, it is also necessary to develop and deploy systems that can accurately detect fraud and report its occurrence to the responsible authority.
Wireless mobile data networks will also suffer from attacks by computer hackers once these networks become widely deployed. The act of using a foreign agent to act on behalf of mobile hosts requires opening the access network to strangers, i.e., the access network is not a tightly administered or restricted network. It is thus essential to have the ability to detect when resources in the access network are being attacked. The hosts of the access network will in fact have access to critical information about users that rely on these hosts to implement agent services on their behalf. The well-developed body of work in host and network intrusion detection might have relevance to this problem.
Various mobility factors such as speed and displacement are characteristics of mobile node's behavior that need to be factored into intrusion detection solutions geared towards mobile and wireless networks.
The specific research topics identified in this area include: