Consumer Applications of Cognitive Radio Defined Networks Sheryl Ball
Adam Ferguson
Thomas W. Rondeau
Dept. of Economics Virginia Tech Blacksburg, Virginia, USA
[email protected]
Dept. of Economics Virginia Tech Blacksburg, Virginia, USA
[email protected]
Dept. of Electrical Engineering Virginia Tech Blacksburg, Virginia, USA
[email protected]
military/emergency services environment from an engineering and business perspective and suggest some strategies for adapting this cognitive radio technology to this environment. We highlight the issues involved in determining what kind of cognitive radio network service to offer to consumers and how to price and allocate the available service among users.
Abstract— Most of the study of implementing cognitive radio technology to date has focused on its potential in military and emergency services uses. This paper examines some of the issues involved in adapting cognitive radio technology consumer markets. We highlight some potential advantages to cognitive radio defined networks for consumers and suggest some possible service designs and pricing systems.
II.
Keywords-cognitive radios, quality of service; physical layer; MAC layer; contingent pricing; dynamic pricing.
A. Military/Emergency Services Use of Cognitive Radio The military and public safety users have direct application for cognitive radios today. With their needs for interoperability and guaranteed quality of service under extreme conditions [1] [2], cognitive radios offer the necessary autonomous operations for these users. Cognitive radios, with their ability to adapt, can adjust to realize the different standards and therefore communicate among and between different standards at the same time. The cognitive radio concept helps in these situations by sensing the environment and learning how to best adjust itself to continue providing a certain quality of service to the public safety radio users.
I.
INTRODUCTION Cognitive radios are a new paradigm in wireless communications that hold promise for new and better services to many markets, including public safety, military, and consumer. Cognitive radios can perceive their environment, learn behavior and environmental patterns, and appropriately adapt themselves to satisfy the immediate needs of the user, network, and radio environment. In order to properly adapt to the present needs, a cognitive radio must be capable of adapting in many directions, altering the waveforms and protocols of the radios without constraint. To date, much of the discussion of cognitive radio has focused on its potential for military and emergency services users. Since both military and emergency situations are difficult to anticipate and dynamic in nature, a radio network that has the ability to design its own optimal and adaptable configuration on demand possesses clear advantages over traditional systems. In addition, cognitive radios are able to operate in and adapt to jamming and interference, ensuring that quality of service is maintained.
Both markets have recently given a lot of attention to the use and development of both software defined radios (SDR), the technical platform to realize cognitive radios, and cognitive radios themselves. The needs in this market are fairly obvious and well-understood, which have been discussed and identified in both our own cognitive radio research group and outside it [3] [4]. What we are interested in now, however, is the extension of these topics into the consumer market. Why does the consumer market differ so much from these markets, and how do we both realize the cognitive radio concepts here as well as understand the economic impact on this, a market more unknown and dynamic than either the military and public safety markets?
Consumer markets offer another potential use of this technology. The inclusion of cognitive radio technology in a consumer network would allow dynamic allocation of physical layer and networking resources that can improve quality and availability of service to consumers and could make feasible business strategies that would improve the financial health of the service provider.
B. How Cognitive Radios Would Fit in a Consumer Market As mentioned above, two of the most important needs of cognitive radios in the military and emergency service markets are for interoperability and QoS guarantees. The needs of consumers are not much different, although as we will develop, there are added dimensions of complexity to be taken into
This paper will focus on the additional allocation issues created when heterogeneous commercial or individual consumers interact with a cognitive radio defined network. We note the differences between this environment and the Sponsored by the National Science Foundation under grants 9983463, DGE-9987586, and CNS-0519959
1-4244-0013-9/05/$20.00 ©2005 IEEE
APPLYING COGNITIVE RADIO TECHNOLOGY TO CONSUMER MARKETS
518
experience slow service and interruptions that lead to dissatisfaction with the service the hotel is providing.
account. Whereas the military and public safety communications call it interoperability, we can look at this same concept as the ability to provide ubiquitous communications to consumers by allowing different radios the ability to communicate. The importance of quality of service guarantees to consumer markets is undeniable, but the purpose is different. Instead of providing reliable and robust communications as required by military and public safety users, cognitive radios allow consumers to trade resources for desired needs of their applications.
One solution is to restrict access to the network and charge a fee. Fees could be based on user type: fees for conference users might be charged to the conference organizes who would then pass the cost along to attendees through conference registration charges; guests paying discounted room rates might pay a high fee for service whereas those paying in premium rooms might have free network access. While sufficiently high rates would decrease demand for Internet access, there is still a potential for quality of service disruptions.
The inclusion of cognitive radio technology in a network allows dynamic allocation of physical layer and networking resources that can improve quality of service, availability of service and the financial health of the service provider and network user. An unsolved problem is that of determining what kind of cognitive radio network service to offer to consumers and how to price and allocate the available service among users.
Without a system of prioritizing users, quality of service problems will continue. Conference participants will be frustrated if a presenter who is trying to conduct a real time demonstration by way of the Internet is experiencing quality degradation in her network access. To facilitate satisfaction with the quality of their conference services, the hotel needs to have a way of making certain that the presenter has access to resources that are needed to conduct a smooth presentation. A crude solution would be to establish two networks, one of which is restricted two a small number of presenters and priority users.
Cognitive radio networks are able to supply a variety of different services, for example, voice and data services in combinations that change to suit user needs. A key difference between a military/emergency user and a consumer is the availability of information about individual users’ needs with regard to quality of service issues such as service priority. The situations encountered by military and emergency service personnel can be more easily defined a priori with respect to communication procedure and desired outcomes. While the physical environments themselves may be more chaotic and indeterminate, the cognitive radios are always required to ensure continuous and reliable communications despite hostile or changing environments.
A cognitive radio equipped network represents a more satisfactory solution to the problem even in a case where the Internet service is provided using a single radio frequency. Each user could be assigned a priority for service use based on the hotel’s service goals and the cognitive radio would optimize access to the network so that higher priority users were given priority over lower priority users, even if it meant that some users were bumped off the network until demand for service decreased. A limited number of conference presenters could be given highest priority to ensure that their presentations go smoothly. At the other extreme, non-conference guests paying discounted hotel rates might have reasonable access on their room floor, but only “space available” access in the busy conference area.
The consumer problem is more complicated because the consumer’s needs at any point in time are known only to the consumer, and thus cannot be programmed into the cognitive engine in advance. When the demands on the network require that some individuals be bumped off or that their quality of service needs to be reduced, the strength of individual’s preferences needs to be known in order to make efficient choices about service allocations. While we predict the environments will be more stable than with the military of emergency services, understanding the user’s needs and adapting the radio to those needs is an added level of complexity not yet analyzed.
III.
ISSUES IN CONSUMER SERVICE DESIGN
A. Spectrum as a Heterogeneous Commodity Unlike the markets for “widgets”, the meaningless homogeneous commodity commonly studied in elementary economics courses, the “good” in a wireless spectrum market has an interesting set of multidimensional characteristics that are important to spectrum consumers. The heterogeneous nature of spectrum allows consumers to make tradeoffs in characteristics in order to maximize their consumption of characteristics that are most important. When aggregated across all users, this system of tradeoff allows for a more efficient utilization of spectrum then is achieved under systems such as First-In/First-Out that treat spectrum as a homogenous commodity. Identifying these characteristics is central to understanding in what dimensions dynamic markets should ultimately be defined. Some characteristics that will be of immediate concern are quality of service (QOS), location, and time of day.
C. A Consumer Service Example An example of a network that provides consumer services and where cognitive radio could be used to provide better services is a conference hotel. Suppose that the conference hotel establishes a traditional unrestricted 802.11 network where the cost of using the wireless network is bundled into the price of conference services and guest rooms. During a large conference, the network will experience demand from presenters at the conference, attendees staying at the conference hotel, attendees staying at the cheaper hotel down the block, and hotel guests traveling on other business or for leisure purposes. Without restrictions on use, anyone in the building with an 802.11 compatible device has equal ability to access the Internet. In times of heavy demand all users may
Quality of service is, however, a somewhat vague characteristic as what is of central concern to one consumer
519
the users must be reachable at all times. This kind of QoS is what cognitive radio research has been strongly focusing on: adaptations to the physical layer to ensure a communications channel with certain guarantees for bandwidth use and data throughput. For these situations we can define a number of objective functions:
might be of little importance to another consumer. An example of this is a consumer who uses spectrum primarily for the purposes of streaming multimedia as opposed to another who uses spectrum primarily for web-browsing. While the user who streams multimedia is primarily concerned with latency and jitter which can cause breaks in the stream, the user who web-browses is not as concerned with latency and jitter.
• • • • • • • •
Wireless spectrum in consumer markets is currently treated as a homogenous product where everyone pays roughly the same fee to receive roughly equal service in the network. Packet sorting and transmission is executed at the queue on a First-in/First-out basis. From an economic standpoint this is not a desirable sorting mechanism as some users may have a greater “need” for higher throughput at a given point in time, as has been pointed out in engineering literature. DaSilva [5] reviews work done in quality of service enabled networks not employing cognitive radio technologies.
bandwidth use, spectrum efficiency, bit error rate, probability of dropped call, throughput, goodput (as opposed to throughput), power consumption, and system delay / computational complexity
The work on these objective functions has dealt greatly with how to manipulate the physical layer parameters including transmitter power, spreading type and code, modulation type and order, pulse shaping, symbol rate, carrier frequency, dynamic range, equalization, and antenna directivity [11] [12] [13]. Other parameters of interest to the cognitive radio adaptation are channel coding, interleaving, and frame size.
B. Lessons from Ipv6 The proposal in this paper is to provide a service and pricing scheme for wireless networks based on the quality of service provided to the consumer. This concept has been studied for a few years in the networking literature for IPv6 (Internet Protocol, version 6), and so it is instructive to study the parallels and differences here.
Without providing the rigorous mathematical background to these communications concepts [14], we argue that the objectives are functions of many of the physical layer parameters; more importantly, the objectives are also competitive: that is, optimization in one objective often hurts other objectives (such as improvement in almost all of the objectives results in poorer performance for the power consumption objective). In keeping with the radio domain vocabulary, we have termed the objective functions “meters” and the radio parameters as “knobs” as we adjust knobs in a radio to alter the performance as read through meters.
IPv6 is the updated protocol to serve the Internet with larger address space, better routing capabilities, and provisions for quality of service guarantees, among others [6]. The address space is taken care of by increasing the length of the address from 32 bits to 128 bits and routing is improved though less complex headers and address hierarchy [7]. QoS concepts were left as an open research problem in the request for comments (RFC) papers by the IETF that define the IPv6 protocol [7]; however, the designers had the foresight to recognize the need for QoS provisioning in the new Internet with the increase in multimedia and other QoS-dependant applications, and so they created “flow label” and “traffic class” parameters as part of the IPv6 packet header. Since then, a lot of work has gone into defining how to set these values [8] and how to charge for different levels of QoS [9] [10].
Quality of service to the network layer, such as was mentioned in the IPv6 discussion, only gives so much of the picture, especially in wireless networks. To radios, providing a guaranteed QoS is not as simple as using a guaranteed bandwidth. Access to the spectrum is a large concern to these systems as well as the tradeoffs in complexity and performance they can achieve. For example, if a user is trying to talk using a simple 64 kbps voice coding strategy, this is equivalent (sans clever source coding techniques on top of the encoding system) to the bandwidth required. In wireless networks, we could say this is equivalent to requiring 64 kHz of spectrum (again, ignoring other factors of the systems and assuming that one symbol per second takes one Hz). However, as part of the physical layer of radios and the intelligence and flexibility of cognitive radios, a modulation scheme could be used that provides 4 bits per symbol such as QAM16. Now, the radio need only transmit at 16 ksps, which means the signal now fits within 16 kHz with an equivalent bit rate of 64 kbps.
The analysis for pricing based on network layer QoS provisioning comes down to a mapping of the bandwidth required versus the total bandwidth available from the network and the importance to the user. This work is both useful, valid, and important to our work on wireless networks, but it does not directly apply to the added complexity of wireless networks, nor does it fully take the concepts to a level for actual implementation – that is, what kind of user interface or user interaction is required in a wireless network for users to indicate their needed QoS and their willingness to pay? C. Cognitive Radio QoS Military and public safety personnel have more welldefined missions and needs in their tasks than consumers do, so their requirements out of a cognitive radio would be more concerned with securing reliable and robust communication channels as well as insuring proper interoperability. These radios would provide some measure of QoS to the extent that
This example is just one way in which the flexibility of the cognitive radios both complicates the analysis but improves the benefits to the system utilization. There are many dimensions of the radio’s physical layer that are tunable and provide tradeoffs in the optimization functions. We can trade power and complexity for bit rate, or add redundancy in the form of an
520
its required quality of service. Adding the analysis of the MAC layer to the network resource sharing problem adds more dependencies to the objective function calculations.
error checking and correcting code to improve reliability while reducing power output. Because cognitive radios allow so much flexibility in their handling of the physical layer parameters, we can realize a lot more dimensions to the quality of service analysis. Power consumption, interference, spectrum efficiency, data rate, error rates, and system delay (latency) among others must all enter into the optimization process to realize the user’s true needs. On the other hand, what our argument here presents is a way to implement a pricing strategy for QoS-controlled network access. To the infrastructure, the pricing scheme does not reflect individual concerns about power consumption or how it will guarantee latency; the system is only concerned with its available resources and the ability to provide those resources to its consumers.
While the cognitive radio uses the physical layer for internal adaptation for the consumer, the MAC layer parameters help realize the infrastructure makeup and can be controlled from the network controller. The cognitive radio network must now find a compromise between not only its own needs, but the available resources provided by the infrastructure. By purchasing the set of resources from the network controller, the cognitive radio now has some context to work within. Using all of the physical and MAC layer techniques available to it, the cognitive radio can trade off all of the performance parameters in order to provide the required QoS to the user as efficiently as possible. That is the first level of analysis; the second level takes the concept further and allows the cognitive radio to negotiate better prices for service. If a cognitive radio thinks it can operate with an equivalent QoS but use fewer resources, the cognitive radio can renegotiate the deal with the network controller for a cheaper price.
So, while we have identified a number of QoS objectives the cognitive radio must solve internally, we now must identify the parameters that affect the QoS that concern the infrastructure. The first and foremost resource available to a network infrastructure is bandwidth; in either wired or wireless networks, bandwidth, while different, produces the same effect: more information. The bandwidth resource has been the focal point of much of the QoS research on the networking level to date. To these analyses, data rate and latency can pretty much boil down to an effective bandwidth that controls how much information can pass in a certain amount of time.
The cognitive radio’s responsibilities are to the user, and so it must act as the agent to provide the best service for the best price. These are decisions that no one should be expected to make every time they want to check their mail, but they are decisions that, when made properly, improve the performance, QoS, and ultimately, satisfaction. However, as we pointed out previously, satisfaction occurs on many more levels than the throughput. System delay and complexity are issues that might cause unnecessary and unwanted delays as well as prematurely drain battery life. A cognitive radio will take these into account to allocate resources that require less complex waveforms for the same QoS. Another radio might not have battery life or delay as problems, and so they can pay less for less specific resource requirements and then use more complex physical and MAC layer parameters to realize their needed level of QoS, such as more robust channel coding or other techniques like spread spectrum.
Other resources available for wireless infrastructure control are time, power, and space. The network controller can provide guaranteed access to radios at certain times, certain maximum power levels, and separate users spatially, either through a cellular concept for spectrum reuse or through antenna beamforming. These techniques do not exist on the network layer, but instead are MAC layer concepts. It is at this point that we can push the research in cognitive radios up the protocol stack from the physical layer into the medium access control (MAC) layer / data link layer (DLL) (we will refer to this simply as the MAC layer and not concern ourselves with the semantic differences between MAC and DLL) [15]. The Medium Access Control layer’s purpose is to share broadcast resources by controlling the method of using channel resources efficiently. Access techniques allow resource sharing in many dimensions such as time, frequency, code, and space [14]. The physical layer parameters already start to play into this concept: choosing a center frequency to minimize interference is a form of frequency division multiple access (FDMA), changing the spreading code could be used for code division multiple access (CDMA), and using beam-forming techniques to alter the antenna directivity is one method of implementing spatial division multiple access (SDMA). However, the analysis on the physical layer for these properties is removed from the multiple access considerations to a large degree, and it provides no means of realizing time division multiple access (TDMA). Whereas the physical layer optimization of the objective functions is more inwardly concerned with the current radios performance, the MAC layer optimization recognizes the needs of other possible radio users to share the resources, which is a consideration when the cognitive radio tries to decide on the optimal settings to achieve
D. Consumer Service Design and Interface Issues The type of service offered to consumers will depend on the environment in which it is established and will likely evolve along with the availability of wireless devices. It seems likely that initial systems will be rather simple, such as the conference hotel example that offers only services for users who wish to access the Internet on a single frequency. As described above, however, one of the strengths of cognitive radio technology is that it can offer users a variety of voice and data services and transmit on a number of different frequencies as needed. In order to utilize a cognitive radio network, consumer devices will need to be modified so that they send identifying information so that the cognitive engine will be able to correctly handle their service needs. The devices will also need to be updated so that they can transmit and receive on multiple frequencies, depending on the available resources. An additional complication in adapting cognitive radio for consumer use is that most consumers are not communications experts and may not be able to articulate their quality of service
521
needs in as much detail as the cognitive engine requires. Even the expert users would likely articulate their needs as a range. For example, how much bandwidth does one need to check one’s e-mail during a break from conference sessions?
workable allocation of wireless resources to users is maximized, regardless of whether all users are permitted to access the network. Efficient pricing mechanisms have been studied in a number of communications environments. [17] [18] [19] [20].
At some point, in fact, the costs to the consumer of accurately articulating their needs may be outweighed by slight improvements in quality of service satisfaction caused by the increased accuracy.
In order for the cognitive engine to come up with efficient solutions, it needs reliable information about user preferences. For example, the cognitive engine needs information about which users really value fast service. If consumers are asked what they want and there are no costs to more resource intensive service, all consumers will say that they want the most resource intensive service. Setting higher prices for more resource intensive services makes it costly for the user to misrepresent their spectrum needs while also automatically prioritizing users based on their willingness to pay for different quality of services. A pricing system must be designed that makes sensible tradeoffs between collecting enough information to be useful and keeping the pricing system simple enough for consumers to make decisions that accurately reveal their preferences.
Research needs to be done to design a workable user interface. There are several criteria by which the interface should be judged a) Whether consumers can successfully articulate their service quality needs b) User friendliness c) Ability to transmit information to the cognitive engine One possible solution is to provide consumers with an ability to select a level of each quality of service indicator, perhaps by means of a slider bar. Consumers might be given a means to access some hints about what level they want to choose of each quality of service indicator, for example, when they click on the slider bar for “bandwidth” a balloon could pop up and indicate whether that level was good for “email” or “Internet browsing” or some other service a consumer might require. Another slider bar might indicate “power consumption” and include recommendations for computers operating on battery power. The complication is that choices on many of the quality of service variables affect feasible choices of others. This means that consumers will either have to correct their original choices, a situation likely to cause frustration, or else the network will have to modify their choices to make them feasible.
We propose three criteria for evaluating a pricing system: 1) Accurately reveals enough information about consumer preferences so that cognitive engine can produce an efficient solution. 2) User friendliness. 3) Practical to administer for involved service provider. The last consideration is critically important, but easily overlooked in a theoretical discussion. While more sophisticated pricing systems that bill users based on time and level of actual use are likely to be most efficient, the business costs of these billing systems will also be greater than less sophisticated pricing. The eventual decision about how to price service must take these tradeoffs into account.
Another solution would be a series of radio buttons where users could indicate what sort of service they plan to use, e.g. e-mail, voice, video, etc. This is likely the user-friendliest solution to the problem; however, it also provides the coarsest information to the adaptation strategies of the cognitive radio and the network. The cognitive radio and the network provider would have to jointly establish the levels of each quality of service variable attached to the choice associated with each radio button.
Figure 1 is a system block diagram of a cognitive engine. The pricing mechanism that we would develop would be integrated as part of the user domain and modeling system.
An “expert” screen might supplement either of the above solutions. This would allow users who have specific needs to override the standard interface and enter information about their precise needs. IV.
COGNITIVE RADIO SERVICE DESIGN AND PRICING STRATEGIES
According to economists, one criterion for evaluating whether the outcome of an allocation process is desirable is whether or not it is efficient. Agents are assumed to have well-defined utility functions that can be used to determine the strength of their preferences for different outcomes. Resource allocations are efficient if they are feasible given the available resources and maximize the sum of the agents’ utilities. In the current discussion, this means that total satisfaction with a
Figure 1. Cognitive Engine Block Diagram.
522
Below we discuss some pricing schemes and begin evaluating their appropriateness for use in a cognitive radio defined network in light of the criteria offered above. We consider three broad categories of pricing systems. The first are fixed fee systems, where prices for service are set in advance and do not vary according to the load on the network. Next we discuss contingent fee systems, where fees may vary with location and time of day based on expected number of users. Finally we consider dynamic price systems, where current information is used to set current prices.
common types of access, users might see the price per increment of time that they would be charged for access.
A. Fixed Fee Systems The Internet was originally built around a single class of user service. All information (in the form of packets) that moved through the network was serviced by routers and switches on a First-in/First-out basis. Users paid a fixed fee for access to the Internet and no additional cost based on their level of use. This relative equality among all users means that the single-service class Internet share features of a public good. As in the case of a public highway, another public good, all users can have as much or as little access to the internet as they wish, or as allowed by their Internet Service Provider. Wireless services share the same public goods feature, whether all users ultimately end up accessing the Internet, or are users of another type of radio services.
A somewhat more efficient solution would allow users to contract in advance for some combination of quality of service and priority on the system. Users who purchase higher maximum quality of service levels or higher priorities would pay correspondingly higher prices. A business user might pay a premium price for “platinum” level service that made them unlikely to be denied access, while a student who had the ability to defer their use to another time when the network was busy could purchase a much cheaper “copper” service level. Among fixed fee systems, consumers might find this approach to be most user friendly and it would also be relatively easy to administer for the service provider. Efficiency gains occur with this type of service because it generates at least coarse information on both quality needs and strength of preferences.
As with all public goods, these systems can create a “Tragedy of the Commons,” where all users have the incentive to use the Internet as much as possible with no regard for the usage needs of their neighbors. This is easily understandable on an intuitive level: users have the incentive to increase their use of service as long as they derive additional benefit from increased usage since there is no cost to doing so. This sort of user behavior results in overuse of the network relative to user valuation and may result in decreased quality of serviced caused by congestion.
B. Contingent Pricing Schemes An improvement over a fixed fee system is a contingent pricing scheme where prices for service with different characteristics is determined in advance and users are charged for the service they select. An example of this type of pricing is electricity pricing that is higher during peak use times of day and lower in the evenings. These systems improve efficiency in electricity service provision by encouraging consumers to defer elective uses, such as running dishwashers, to non-peak times of day. This reduces the load on the system and increases satisfaction with the system for producers, who can avoid additional investment in capacity, as well as consumers, who save money with respect to flat rate pricing.
There are further advantages to taking this approach. As long as users have regularly shaped demand for service, this would decrease demand for service. With sufficiently high prices, this system would reduce demand on the system enough to decrease the frequency with which congestion problems occur. What this system fails to do is generate information on which users have the highest utility for service, so inefficiencies caused by denying the wrong user access to service still occurs.
Suppose that cognitive radio technology is used in conjunction with a fixed fee system where users are charged a fixed fee, known (and perhaps paid) in advance, for access to the system. In this case, the network would refuse to accept additional connections once its capacity is reached. This would provide limited improvement over current technology. While the system could be designed so that everyone allowed to access the network has good quality service by bumping some people off, you are not necessarily doing a good job of using spectrum efficiency. For example, you may be allowing a person to download a recipe they do not need for several days while at the same time refusing service person sending details of a time sensitive and high value business matter. In order for the system to work efficiently, you need to know who values the service the most and allocate the service to them at the expense of those who value service the least.
Our problem is substantially more complicated than the electric power example, however. In the conference hotel example discussed above, prices would need to be presented to users as a matrix of different service characteristics at different times and possibly in different locations within a facility. For example, it might be expensive to send e-mail during times when the conference is offering on the conference floor of the hotel, but inexpensive to use the service at the same time on a guestroom floor or when the conference is not in session. The complexity of the pricing matrix might make the decision about whether or not to use service prohibitively complex for the user. This “user unfriendliness” can result in potential users being averse to trying the system as well as decreased satisfaction among those who do eventually become users. In evaluating this type of contingent pricing scheme, we will examine possible variations that will make our proposal easier for consumers to understand.
A more sophisticated system that takes better advantage of the benefits of cognitive radio technology would integrate pricing with the level of service offered to users. When a user first tries to access the system they might encounter an introduction screen where they enter some information about the type of service that they need (see discussion above) before accessing the system. In the case where users choose between
This may be an additional argument for offering users a restricted menu of services based on pre-selected quality of
523
value of service being displaced by a user when network traffic is high. While this means that the network is “locked in” to providing service to this user for the promised amount of time, the risk of bumping off a user with a higher willingness to pay for service might be outweighed by the relative simplicity of the system.
service characteristics. Users could further restrict the parameters by time of day and location. Ultimately, this would allow them to generate a more manageable menu of choices that still allows them to make good decisions. A conference attendee from our example above might indicate that they need to check e-mail by the end of the day and generate a short list of choices that let the decide whether to pay a premium to check their mail on the conference floor or to return to their room to pay a lower price.
D. Evaluating Pricing Systems A number of approaches will be needed in order to adequately evaluate any type of pricing system. Each candidate pricing system needs to be formalized so that it is clear what information needs to be passed to potential users and what information is available for the cognitive engine at each point in time. Then candidate systems should be evaluated theoretically so that their efficiency characteristics are well understood.
This type of contingent pricing system is still choosing prices in advance based on historical use rates. As such, its ability to generate efficient solutions is limited as is the ability to provide the cognitive engine with information about the strength of user preferences. C. Dynamic Pricing Schemes Vickrey [16] discusses contingent pricing systems and argues that features of the service being offered make the timing of price setting critical to user acceptance of the pricing system. Because electric power consumption is relatively predictable it is efficient to announce prices in advance. As discussed above, consumer users can defer some electricity consumption if the cost savings justifies the delay. On the other hand, he argues that the price of long-distance phone service could be priced at the time a call is initiated by informing the user of the current cost per minute of service via recorded message once the number is dialed. Users who find the current charge too high can hang up and try at a later time. Wireless service in a cognitive radio defined network could operate in a similar way.
Next the interaction of the proposed systems with the cognitive engine should be studied. The specifics of the way that the pricing system will be integrated into the cognitive engine needs to be designed and tested. Next simulations can help determine whether the cognitive engine is able to produce acceptable solutions in a reasonable period of time. Finally, experiments using a cognitive radio testbed and likely user demand characteristics will allow additional fine-tuning of the system. Another important step in evaluating the system is to study how potential users interact with the system. This can be accomplished through experiments with human subjects. Again, there are a number of types of studies that need to be conducted. One group of potential experiments would ask subjects to evaluate the usefulness of the user-interface qualitatively. Subjects should also be asked to make decisions to achieve different use objectives and the accuracy of their choices measured.
All possible pricing schemes can be classified as either static or dynamic pricing or some combination of the two. Static pricing encompasses schemes which price contingent upon service class or time of day usage but is independent of current network utilization. Dynamic pricing schemes, on the other hand, tie the pricing directly into network conditions where the price can fluctuate based on the current network utilization [5]. This makes dynamic schemes much more complex computationally and much more costly since there must be a mechanism for constantly feeding network information into the pricing algorithm. The cognitive engine will be an advantage in any attempt to implement this sort of scheme.
Experiments are also needed to study the interaction of human subjects, the pricing system and the cognitive engine. Following market testbed experiments conducted by economists, subjects could be given a known preference profile and asked to make decisions that would be fed into the cognitive engine. Results could be evaluated for efficiency, quality of the solution the cognitive engine generates and time to arrive at the solution.
Users may have a hard time understanding how they will be charged under these complex systems and as a result may find dynamic schemes unappealing. However, dynamic pricing schemes promise to do a better job of providing accurate information about user preferences to the cognitive engine thus facilitating better optimization of available resources.
V.
CONCLUSION
In this work, we have begun a discussion of the research problems to be solved before consumer services can be provided by a cognitive radio defined network. These include: 1. Defining the services to be offered and a user interface
Careful design of a pricing system can eliminate inconvenience to consumers and increase acceptance. As in the above description of defining user service, the user interface design is critical. An example of a promising user interface would create a login system where users select the level of service they wish to use (e.g., Internet browsing) and are quoted a price per unit of time ($0.50 for 15 minutes) that is good for a fixed period of time (1 hour). There would be a minimum price for each level of service, with prices set at the
2. Determining how to price the available service 3. Designing a means to integrate information produced by consumers about their needs into the cognitive engine. Since the solution to each of these questions is integral to each of the others, a positive strategy is for an interdisciplinary team to use an iterative approach.
524
The study of cognitive radios involves the entire problem, from technical challenges of how to create such intelligent and robust radios to understanding the regulatory and policy impacts of these radios. We have deliberately declined to address the latter problem in this work.
[9]
[10]
[11]
ACKNOWLEDGMENT This work was supported by the National Science Foundation under grants 9983463, DGE-9987586, and CNS-0519959.
[12]
REFERENCES [1]
[2] [3]
[4]
[5]
[6] [7] [8]
[13]
The SAFECOM Program, Department of Homeland Security, “Statement of Requirements for Public Safety Wireless Communications and Interoperability,” Version 1.0. Mar. 10, 2004. http://www.safecomprogram.gov/SAFECOM/interoperability/default.ht m “Joint Tactical Radio System (JTRS),” http://jtrs.army.mil/ C. W. Bostian, S. F. Midkiff, T. M. Gallagher, C. J. Rieser, and T. W. Rondeau, "Rapidly Deployable Broadband Communications for Disaster Response," Sixth International Symposium on Advanced Radio Technologies (SAFECOM Session), Boulder, CO, pp. 87-92, 2004. V. T. S. Shi, "Evaluating the performability of tactical communications networks," IEEE Trans. Vehicular Technology, vol. 53, pp. 253 - 260, 2004. L. A. DaSilva, “Pricing for QoS-Enabled Networks: A Survey,” IEEE Communications Surveys, http://www.comsoc.org/livepubs/surveys/public/2q00issue/dasilva.html, Second Quarter 2000, pp. 2-8. H. Huang and J. Ma, "IPv6 - future approval networking," IEEE Proc. WCC - ICCT, pp. 1734 - 1739, 2000. S. Deering and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification," RFC2460, 1998. A. Feldman and S. Muthukrishnan, "Tradeoffs for packet classification," IEEE Proc. INFOCOM, pp. 1193 - 1202, 2000.
[14] [15] [16]
[17]
[18]
[19]
[20]
525
C. A. Courcoubetis, A. Dimakis, and M. I. Reiman, "Providing bandwidth guarantees over a best-effort network: call-admission and pricing," IEEE Proc. INFOCOM, pp. 459 - 467, 2001. J. A. V. Mieghem, "Price and Service Discrimination in Queuing Systems: Incentive Compatibility of Gc|mu Scheduling," Management Science, vol. 46, pp. 1249-1267, 2000. T. W. Rondeau, B. Le, C. J. Rieser, and C. W. Bostian, "Cognitive Radios with Genetic Algorithms: Intelligent Control of Software Defined Radios," Software Defined Radio Forum Technical Conference, Phoenix, AZ., pp. C-3 - C-8, 2004. S. Haykin, "Cognitive Radio: Brain-Empowered Wireless Communications," IEEE Trans. Selected Areas in Communications, vol. 23, pp. 201 - 220, 2005. V. D. Chakravarthy, A. K. Shaw, M. A. Temple, and J. P. Stephens, "Cognitive radio - an adaptive waveform with spectral sharing capability," IEEE Proc. WCNC, pp. 724 - 729, 2005. J. G. Proakis, Digital Communications, 4 ed., New York: McGraw Hill, 2000. A. S. Tanenbaum, Computer Networks, 4 ed., Upper Saddle River, NJ: Prentice Hall, 2003. W. Vickrey, “Responsive Pricing of Public Utility Services,” The Bell Journal of Economics and Management Science, vol. 2, no. 1, Spring 1971, pp. 337-346. F. C. Charmantzis, C. Courcoubetis, V. Siris, and G. D. Stamoulis, "A comparative study of usage-based charging schemes," IEE Colloquium Charging for ATM, pp. 6/1 - 6/7, 1996. R. Cocchi, S. Shenker, D. Estrin, and L. Zheng, “Pricing in Computer Networks: Motivation, Formulation, and Example,” IEEE/ACM Transaction on Networking, vol. 1, No. 6, December 1993, pp. 614-627. E. Viterbo, C. F. Chiasserini, “Dynamic Pricing for Connection-Oriented Services in Wireless Networks,” Personal, Indoor and Mobile Radio Communications, 2001 12th IEEE International Symposium on, vol. 1, 30 Sept.-3 Oct. 2001, pp. A-68 - A-72. S. Yaipairoj, F. C. Harmantzis, “Dynamic Pricing with ‘Alternatives’ for Mobile Networks,” IEEE , vol. 2, 21-25 March 2004 pp. 671 - 676 .