Harvard SEAS Theoretic Fundamentals, Regulatory issues, Physical Limitations, and the Future of Opportunistic Transmission Vahid Tarokh Harvard University UC Irvine-CS Dept Feb 24. 2012 1 Harvard SEAS Introduction UC Irvine-CS Dept Feb 24. 2012 2 Harvard SEAS The Goal The Goal = Providing Wireless Services UC Irvine-CS Dept Feb 24. 2012 3 Harvard SEAS Services •Traditionally voice had been the main application, but many other services are arising. •Emerging wireless broadband applications require both spectrum and advanced techniques to increase bandwidth efficiency. Example of Services Information Services Software Distribution Entertainment Television Interactive Games Education Services Electronic Shopping UC Irvine-CS Dept Feb 24. 2012 4 Harvard SEAS Impact of Services • Historically by providing services, telecommunication engineers have had a huge impact on society and economy. • New services enable new non-telecom industries and improve efficiency of existing ones. • They may help with development of freedom and democracy UC Irvine-CS Dept Feb 24. 2012 5 Harvard SEAS A Successful Example • Newsweek reports that cell-phones have made 5 major impact on the world http://www.newsweek.com/2010/11/10/how-thecell-phone-is-changing-the-world.html – Exposing Secrets - The repression and horror happening in North Korea leaks out by cell phone. – Advancing Democracy - Cell phones present a problem for oppressive regimes everywhere. – Enabling Commerce - Enabling a common method of banking using a cell phone where there are no banks. – Distributing Medecine - A new project in Africa, called Stop StockOuts enables activisits to report which drugs are out of stock. – Waging War - How the Taliban have forced local cell-phone-service providers to shut down their towers at night stopping locals from reporting Taliban movements to Coalition forces. UC Irvine-CS Dept Feb 24. 2012 6 Harvard SEAS Communications Help Economy and Human/Political Development • Kerala to send SMS alerts for vaccination of babies • South African Students Receiving Maths Lessons by Mobile Phones • Kenyan farmers use SMS to beat climate-driven price uncertainty. • Saving Mothers' Lives With Health Tips Via Phone Source: http://www.textually.org/textually/archives/2011/04/028310.htm • Mobile Internet played an important role in Arab Uprising and the Iranian anti-government protests. UC Irvine-CS Dept Feb 24. 2012 7 Harvard SEAS Communications Enhances Freedom • Videos collected by mobile phones expose government and their brutality against citizens. • Providing more services communications can lead to a more “information flat world”, where everyone will access the information that they need and contribute to information gathering and distribution as much as they could. – This will make suppression of the truth much harder. • Thus providing more services is a great idea. – But obviously this needs spectrum UC Irvine-CS Dept Feb 24. 2012 8 Harvard SEAS Spectrum UC Irvine-CS Dept Feb 24. 2012 9 Harvard SEAS Scarcity of Spectrum • Most frequency bands up to 6 GHz (and beyond) have FCC allocations for multiple users. UC Irvine-CS Dept Feb 24. 2012 10 Harvard SEAS Scarcity of Spectrum Source cnn.com Feb 21, 2012 UC Irvine-CS Dept Feb 24. 2012 11 Harvard SEAS Spectrum Crunch Source money.cnn.com Feb 21, 2012 UC Irvine-CS Dept Feb 24. 2012 12 Harvard SEAS Outline • Shortage of good spectrum may appear as a problem, but this may not be the case: – measurements show that at anytime more than 90% of these resources are not used. • Idea: Intelligent radios may allow better use (sharing) of the spectrum. UC Irvine-CS Dept Feb 24. 2012 13 Harvard SEAS Spectrum Sharing • This motivated a push for sharing the unused but dedicated spectrum for providing new services. – Is this a new idea? • Geographical reuse of spectrum has been around for a long time. – – – – – – Is this a good idea? How aggressively must it be pursued/allowed? Is it technically feasible? How much intelligence is needed in the radio? How regulatory bodies are dealing with it? Does it have a future? UC Irvine-CS Dept Feb 24. 2012 14 Harvard SEAS Existing Spectrum Sharing Examples • Spectrum Sharing is nothing new: – The ISM band allows sharing of spectrum • Many successful application exists (e.g. garage door openers, etc.) • Perhaps the most successful application is WiFi. – UWB • A February 14, 2002 Report and Order by the FCC authorizes the unlicensed use of UWB in the range of 3.1 to 10.6 GHz. The FCC power spectral density emission limit for UWB emitters operating in the UWB band is -41.3 dBm/MHz. • This is the same limit that applies to unintentional emitters in the UWB band, the so called Part 15 limit. However, the emission limit for UWB emitters can be significantly lower (as low as -75 dBm/MHz) in other segments of the spectrum. – UWB has not had much commercial success yet. UC Irvine-CS Dept Feb 24. 2012 15 Harvard SEAS Wi-Fi • Another example of wireless broadband services is Wi-Fi. •Wi-Fi has had enormous success and this is expected to continue on into future: please see the old forecast • Range is an issue. UC Irvine-CS Dept Feb 24. 2012 16 Harvard SEAS Existing Spectrum Sharing Examples • MVDDS (Multichannel Video and Data Distribution Service) – This terrestrial based wireless transmission method reuses Direct Broadcast Satellite (DBS) frequencies for distribution of multichannel video and data over large distances. – licensed for use in the United States by the FCC. • Ruled that 10% increase in rain outages would not be harmful – The underlying spectrum is in the 12.2 - 12.7 GHz range. UC Irvine-CS Dept Feb 24. 2012 17 Harvard SEAS Spectrum Sharing Paradigms • Exclusive access: – One system has exclusive access to the spectrum. • Horizontal Sharing (Equal right access): – All systems have the same regulatory status and may access the spectrum on an equal footing. (e.g. usage of the ISM bands by WLAN and Bluetooth) • Vertical Sharing (Prioritized spectrum access): – A primary system. – Secondary systems can share only if they do not generate harmful interference for the primary • All of these schemes have existed for some time. UC Irvine-CS Dept Feb 24. 2012 18 Harvard SEAS Spectrum Sharing Models Spectrum sharing Spectrum sharing with other (legacy or other novel) systems. Pre-established priorities for all involved systems. Horizontal sharing with coordination All involved systems share the spectrum based on a set of rules (spectrum etiquette). Horizontal sharing without coordination Little can be done To avoid interference UC Irvine-CS Dept Feb 24. 2012 Vertical sharing Secondary system has to control its emissions to prevent interference towards primary system. 19 Harvard SEAS Cognitive Radios? •Then What is so new about cognitive radios? •Perhaps it all depends on what cognitive radio means…. UC Irvine-CS Dept Feb 24. 2012 20 Harvard SEAS Cognitive Radios UC Irvine-CS Dept Feb 24. 2012 21 Harvard SEAS Cognitive Radios • A Cognitive radio is an intelligent wireless communication system that: –is aware of its environment, –learns from the environment, and – adapts its internal states in real-time UC Irvine-CS Dept Feb 24. 2012 22 Harvard SEAS Vertical Sharing • Following problems need to be addressed: – need to identify the spectral “white spaces’’ – need to adapt to the restrictions identified. – need to be smart in reducing the harmful interference to other systems while increasing their own transmission rates • Cognitive Radios are being discussed from several perspectives. • Their success will depend on: – Fundamental Limits – Finding methods to achieve these limits – How conservative the regulatory rules are, and – Economics and business models UC Irvine-CS Dept Feb 24. 2012 23 Harvard SEAS Fundamental Limits on Cognitive Radios UC Irvine-CS Dept Feb 24. 2012 24 Harvard SEAS Scenario: Cognitive Radio [DMT] Traditional Cognitive Radios UC Irvine-CS Dept Feb 24. 2012 25 Harvard SEAS Potentials of Cognitive Radios [DMT] Can these potentials be actually realized? UC Irvine-CS Dept Feb 24. 2012 26 Harvard SEAS Interference Between Cognitive Devices • The previous result gives some promise in potential of secondary users not having harmful effects on the primary user capacity. • However, even if we can address the issue of interference between primary and cognitive networks, we will still have potential interference between various cognitive networks operating on available white spaces. • In other words, if because of availability of free spectrum many secondary systems emerge, can these systems support any reasonable data rate? UC Irvine-CS Dept Feb 24. 2012 27 Harvard SEAS Throughput Scaling • From point of view of scaling laws (growth order), of ad hoc networks: “cognitive networks achieve throughput scaling of a homogeneous network”, [WDVCT]: – Randomly distributed n primary users, and m secondary users with m = nβ with β > 1 – Specifically, the primary network achieves the sum throughput of order n0.5 and, for any δ , the secondary network achieves the sum throughput of order m0.5- δ with an arbitrarily small fraction of outage. – These results are only of theoretical interest. UC Irvine-CS Dept Feb 24. 2012 28 Harvard SEAS Co-existence of Secondary Networks • It has been proved [VT] that under the assumption of a cap on the interference caused by secondary network to primary receivers – the secondary networks are single hop and – transmissions transmit either (i) with constant transmit power, and (ii) with transmit power scaled according to the distance to a designated primary transmitter, then as the number of secondary networks N ∞ , the secondary receivers can achieve at least a non-vanishing throughput. • This shows that cognitive radios are at least scalable for single hop networks. • Another option is not to allow too many secondary networks. UC Irvine-CS Dept Feb 24. 2012 29 Harvard SEAS Co-existence of Secondary Networks • The FCC proposes that some form of contention protocol be employed to reduce the interference between co-existing cognitive networks but does not specify such a protocol. • If the number of cognitive networks in a region grows large, this may not be very efficient and may produce capacity losses. • The general problem of allocation of available white spaces to various cognitive networks in order to optimize the capacity is a. computationally hard (NP-hard) problem. • We will next study proposals for various secondary networks to co-exist. UC Irvine-CS Dept Feb 24. 2012 30 Harvard SEAS Existing Approaches • Methods based on Iterative Water-filling (IW): – – • Methods based on graph coloring: – – – • High computational complexity. Convergence to configurations which are far from optimal. Computationally expensive Too much message passing among the agents. Complex cooperation protocols. A Method (GADIA) inspired by Glauber Dynamics in statistical physics [BT] that attempt to maximize the Rosenthal potential. UC Irvine-CS Dept Feb 24. 2012 31 Harvard SEAS Performance • The GADIA algorithm converges to equilibrium exponentially fast. • The GADIA algorithm achieves about 98% of the optimal Shannon capacity. • GADIA has much lower complexity (about 3 orders of magnitude lower) and converges faster that the existing Iterative Water-filling algorithm. UC Irvine-CS Dept Feb 24. 2012 32 Harvard SEAS Conclusions • Theoretical Analysis indicates that under idealized assumptions at least for some scenarios of interest, cognitive radios may have some promise. – There is a lot more to investigate particularly if the idealized assumptions are removed • The main question is that how much of these gains remain – in realistic situations, and – under the regulatory restrictions. UC Irvine-CS Dept Feb 24. 2012 33 Harvard SEAS Regulatory Issues UC Irvine-CS Dept Feb 24. 2012 34 Harvard SEAS FCC Regulatory Issues • The FCC has released the band 3650-3700 MHz for cognitive transmission. • Fixed Satellite Services and federal government stations are currently transmitting in this band. • Certain geographical areas around these transmitters are not allowed for secondary transmission. • Otherwise secondary transmissions are allowed (by the FCC) subject to – 25W per 25 MHz bandwidth for fixed stations – 1W per 25 MHz bandwidth for mobile stations UC Irvine-CS Dept Feb 24. 2012 35 Harvard SEAS TV White Spaces • Another band of interest is given in notice of rule making ET Docket 04-186. • These are TV Broadcast bands (6 MHz channels designated channels 2 to 69 in the VHF and UHF portions of the radio spectrum. • 54-72 MHz, 76-88 MHz, 174-216 MHz and 470806 MHz. • Other existing devices in some of this band include wireless microphones. UC Irvine-CS Dept Feb 24. 2012 36 Harvard SEAS TV White Spaces • Secondary transmission in this band has witnessed a lot of politics/resistance. • Finally, the FCC has announced on Nov. 4, 2008 a set of rules for secondary devices to operate in TV bands while reducing the interference to primary users. • This has caused some interest in network solutions and consumer devices for these bands. UC Irvine-CS Dept Feb 24. 2012 37 Harvard SEAS FCC Rules • All devices, except personal/portable devices operating in client mode, must include a geo-location capability and provisions to access over the Internet a database of protected radio services and the locations and channels that may be used by the unlicensed devices at each location. • The unlicensed devices must first access the database to obtain a list of the permitted channels before operating. • The database will be established and administered by a third party • The third party (Spectrum Bridge) was selected through an open process to solicit interested parties in 2011. UC Irvine-CS Dept Feb 24. 2012 38 Harvard SEAS FCC Rules • Fixed devices may operate on any channel between 2 and 51, except channels 3, 4 and 37, and subject to a number of conditions such as a restriction against cochannel operation or operating adjacent to TV channels. • Fixed devices may operate at up to 4 Watts EIRP. • Personal portable devices may operate on any unoccupied channel between 21 and 51, except channel 37. • Personal portable devices may operate at up to 100 mW of power, except that operation on adjacent channels will be limited to 40 mW. UC Irvine-CS Dept Feb 24. 2012 39 Harvard SEAS FCC Rules • Fixed and personal/portable devices must also have a capability to sense TV broadcasting and wireless microphone signals as a further means to minimize potential interference. • Wireless microphones will be protected in a variety of ways. The locations where wireless microphones are used, such as entertainment venues and for sporting events, can be registered in the database and will be protected as for other services. In addition, channels from 2 – 20 will be restricted to fixed devices. • In addition, in 13 major markets where certain channels between 14 and 20 are used for land mobile operations, channels between 21 and 51 are left free of new unlicensed devices. UC Irvine-CS Dept Feb 24. 2012 40 Harvard SEAS FCC Rules • All fixed devices must register their locations in the database. • In addition, fixed devices must transmit identifying information to make it easier to identify them if they are found to interfere. Furthermore, fixed and personal/portable devices operating independently must provide identifying information to the TV bands database. • All devices must include power control so that they use the minimum power necessary to accomplish communications. • All white space equipment must be certified by the FCC Laboratory. • FCC permits applications for certification of devices that do not include the geo-location and database access capabilities, and instead rely on spectrum sensing to avoid causing harmful interference, subject to a much more rigorous set of tests by the FCC Laboratory. UC Irvine-CS Dept Feb 24. 2012 41 Harvard SEAS Conclusions • A fixed device must employ both geo-location, database access and spectrum sensing capabilities that enable the device to listen for and identify the presence of signals from other transmitters. • A personal/portable device must either be under the control of a fixed device or a personal/portable device that employs geo-location, database access and spectrum sensing or employ geo location/database access and spectrum sensing itself. • These devices will be required to sense, at levels >= -114 dBm, signals of other services. UC Irvine-CS Dept Feb 24. 2012 42 Harvard SEAS Assessing The Rules UC Irvine-CS Dept Feb 24. 2012 43 Harvard SEAS Summary • For reducing interference the FCC proposed methods are based on – Transmit power limitations/power control – Geo-location enabled devices – Geographic databases – Career sensing – Beacon detection – Combinations of these methods UC Irvine-CS Dept Feb 24. 2012 44 Harvard SEAS Geo-Location Enabled Devices • In this method of interference reduction, secondary users must be endowed by GPS (or similar geo-location systems) with at least 300m accuracy. • Primary users location is known to the secondary users (using a geographic database) and buffer regions around the primary users are specified where secondary user transmissions are not allowed in certain bands. • Geo-location and also FCC power limits are safe but conservative: – May make more sense to allow different power limits in various bands based on the location of the secondary user. UC Irvine-CS Dept Feb 24. 2012 45 Harvard SEAS Career Sensing • Secondary devices sense the channel and based on the activity level decide if it is busy or not. • FCC: -114 dBm power means the channel is busy. A B Failure Causes Interference Career Sensing C UC Irvine-CS Dept Feb 24. 2012 D 46 Harvard SEAS Career Sensing • It is obvious that -114 dBm is not the optimum threshold for detecting a busy channel. • If this threshold is not correctly selected it limits the efficiency of cognitive devices, thus – Optimum threshold for detection must be computed although: • Typically the underlying ambient noise std б is not known • The distribution of the primary signal is not known. • The busy channel threshold must be selected based on geographic region (and the underlying primary systems) at least for devices using geo-location and databases. • Similar conclusion can be made for beacon detection. • Here we have to be also careful about transmission strategy. UC Irvine-CS Dept Feb 24. 2012 47 Harvard SEAS Assumptions • Assumptions: • Ambient noise is Gaussian with zero mean and an estimate of б can be obtained • Primary transmission power is Pp. • For career sensing, the primary signal is Gaussian with mean zero and variance Pp. – These are reasonable assumptions if each secondary user scans and averages the channel for some reasonable time (during both idle and busy periods). UC Irvine-CS Dept Feb 24. 2012 48 Harvard SEAS Issues • Computation of CCT (Clear channel threshold) for deciding on idle channels and associated detection strategies for both sensing and beacon based systems is a straightforward exercise in detection theory. • Questions: – Given a peak power Ps and average transmit power Pav for the secondary user, what is the best secondary user transmission strategy that minimizes the interference to primary receivers? – The answer is non-trivial and is given by the following theorem. UC Irvine-CS Dept Feb 24. 2012 49 Harvard SEAS Results • Theorem [KGMT]: – The best transmit strategy (one that minimizes interference to primary users for a fixed average and peak transmission power) for career sensing based or beacon detection based cognitive radios are identical: – The cognitive radio must transmit at full power Ps when detection reliability (LLR between clear channel and busy channel hypotheses) is above a certain transmission threshold (TT) [different than CCT] and refrain from transmission otherwise. This TT depends on • detection being beacon based or career sensing based • average transmit power Pav • Average idle time of primary transmitter. • Thus even when CCT is correctly set based on location, one should operate based on TT (again set based on location). UC Irvine-CS Dept Feb 24. 2012 50 Harvard SEAS Conclusions • Given the discussion, it seems that FCC rules could be made less conservative. – Not surprising given that the FCC is a political organization. – FCC leaders are appointed mostly for their political ties and are not usually from the engineering community UC Irvine-CS Dept Feb 24. 2012 51 Harvard SEAS Economics and Business Models UC Irvine-CS Dept Feb 24. 2012 52 Harvard SEAS Economist Views • There are various economist views about cognitive radios – Some view spectrum sharing as less beneficial in a long run than exclusive model. [please see paper by Coleman Bazelon, Brattle Group in DYSPAN 2008]. – Most economists do not like the idea of sharing – Nevertheless spectrum sharing in some bands has enabled Wi-Fi which has had tremendous success • I am personally optimistic that spectrum sharing (in some bands) is a good idea. UC Irvine-CS Dept Feb 24. 2012 53 Harvard SEAS Services • What kind of services cognitive radios can enable? • The answer is already known in the horizontal sharing scenario. • For vertical sharing scenarios if there is an active primary user in the area, then quality of service may be an issue, unless secondary signals can be spatially separated from that of primary signals. • If multiple cognitive radios exist, then contention can effect their ability to provide quality of service. • Possibly some polling of dedicated spectrum (wireless or wired) with cognitive radio spectrum can be used to provide some quality of service by future service providers. UC Irvine-CS Dept Feb 24. 2012 54 Harvard SEAS Conclusions • Future services that will emerge on cognitive radio spectrum will either – separate their signals from primary signals spatially, or – must be tolerant to delays and lack of quality of service in the event of active primary systems, – or must poll their cognitive spectrum with dedicated spectrum/resources (wireless or wired) in order to provide quality of service. • It remain to be seen what kind of fundamentally new services can emerge (given these constraints) that otherwise will not be possible. UC Irvine-CS Dept Feb 24. 2012 55 Harvard SEAS Envisioned Radios • At present standardization efforts are under way for cognitive Radios. • These systems will be typically used in rural areas where there is not a primary user. • In England, BT wants to provide Internet services in rural areas to consumers given that white spaces exist. • Similar efforts exist in USA. • However, the underlying radios are not truly cognitive. UC Irvine-CS Dept Feb 24. 2012 56 Harvard SEAS The Future of Cognitive Radios UC Irvine-CS Dept Feb 24. 2012 57 Harvard SEAS Final Thoughts • Some promising theoretical gains exist if the wireless devices become intelligent. – It remains to be seen if these gains can be realized given realistic constraints. – Co-existence limitations between cognitive networks is not fully understood yet-- although some results exist. • Conservative regulations (by regulatory bodies) may not allow for a fully cognitive radio to be realized. – However, the FCC allows for sharing of certain bands subject to some etiquettes and rules (a concept that existed before) • Quality of service is hard to obtain unless cognitive radios poll their secondary resources with other dedicated resources (to fall back on). UC Irvine-CS Dept Feb 24. 2012 58 Harvard SEAS Final Thoughts • Will cognitive radios produce services otherwise not easily/economically feasible? • Because some spectrum is now dedicated by the FCC, there will be some devices and services in these dedicated bands – people like free spectrum • The main question is how the communication protocols of these devices and the services will be fundamentally different from protocols existing in the literature. – will these devices be truly disruptive? UC Irvine-CS Dept Feb 24. 2012 59 Harvard SEAS Final Thoughts • As we can see there is a lot of uncertainty about the amount of intelligence a future radio must have. • There is also debates about how sharing must be done. • Much more analysis must be done and many engineering issues must be resolved to answer these questions. • At the end, the devices may end up using the old protocols. • Nevertheless, however the landscape may turn out to be, there is no doubt that intelligent radios remains an intriguing topic. UC Irvine-CS Dept Feb 24. 2012 60 Harvard SEAS Thanks a lot UC Irvine-CS Dept Feb 24. 2012 61