Wireless Communications: The Future Professor Song Chong Network Systems Laboratory EECS, KAIST song@ee.kaist.ac.kr

Current Position A wide range of wireless devices Mobile Fixed Short-range Broadcasting Cellular 2G 3G WiMax So-called 4G Mobile Mesh Emerging technologies: cognitive radio (CR), software-defined radio (SDR)

Current Position Fixed Short-range Broadcasting Point-to-point Point-to-multipoint Fixed mesh Short-range W-LANs 802.11 family Zigbee W-PANs BlueTooth High-speed variants such as WiMedia/UWB RFIDs Broadcasting Analog and digital broadcasting Mobile broadcasting

3G Cellular In 2006, 3G systems were starting to be widely deployed. W-CDMA (European standard), Cdma2000 (US standard), TD-SCDMA (Chinese standard) 3G will eventually take over from 2G but the growth may occur more because the operators ‘push’ the new technology than the subscribers demand it. Realistic data rates of 3G will not go beyond around 400 Kbps. The lifetime of 3G will be around 20 years. The wide use of new services 3G offers such as video call and streaming will take 10 years. As of today, the benefits of 3G were being realized as increased voice call capacity. 3G will face competition from other technologies. W-LAN for hotspot and indoor voice and data WiMax for outdoor voice and data, although this is not certain

4G Cellular The deployment of 4G sometime around 2014-2018 might look like a fairly certain bet. Definition of 4G is still opaque, 4G is likely to be different from 3G (not just be a new air interface), and perhaps may not even emerge.

4G Cellular Each generation has accepted a smaller cell size in return for a higher data rate. Higher data rate -> more spectrum -> higher frequency -> lower propagation range -> smaller cell The next step in the process, where 4G might logically fit, is already taken by a mix of 3G enhancements, WiMax and WiFi. E.g., the Japanese plan for 4G (OFDM, 3-6 GHz band, 100 Mbps) is almost identical to the specification for 802.11a 4G systems, if realized, can be economically deployed only in high-density areas. A further increase in air interface data rate is pointless without better backhaul technologies. E.g., insufficient speed of ADSL There may not be sufficient economic justification for the development of a completely new standard like 4G. Instead, might expect to see novel enhancements to the current standards making up the complete picture. E.g., WiFi-like cellular, cellular-like WiFi, femto-cell network

Prognosis for Cellular A long period of stability, with profitable operation and deployment of 3G, is expected. The likelihood of dramatically new or destabilizing technologies appears to be low. There appear to be few threats to cellular revenues, with the exception of in-building voice calls transferring to W-LAN over time.

Short-range Devices Potential applications for short-range devices are those that are not well suited to cellular. Networking around the office or home High-speed data transfer Cable replacement Machine-to-machine communications

Prognosis for Short-range Devices W-LANs and BlueTooth will dominate the short-range devices market, providing building networks and device-to-device connectivity, respectively. WiMedia/UWB is still a developing technology and it is unclear whether there are sufficient applications that need its very high data rates. Zigbee is likely to succeed as a niche standard for specific applications where widespread interoperability is not needed but battery life is critical. RFID is used in quite different applications from other short-range devices. There is little reason why it cannot continue to be successful.

How People React to New Technologies

How People React to New Technologies A new service or product might take 4 to 10 years to reach mass adoption. Adding 5 years for the standardization and development, it might take 15 years from conception to large-scale success. It is unlikely that total communications spending will grow by more than 0.15% of household income per year.

Spectral Efficiency is Approaching Limit Under some assumptions, Shannon’s law yields where = number of users that can be supported = total available bandwidth = user bit rate = closeness to the Shannon limit For a system with =1, = 1MHz and = 10 Kbps, = 142 calls. 3G systems with HSDPA enhancement attain 30-50 calls per cell, delivering about a third of the maximum capacity achievable. Efficiency of wireless systems is approaching fundamental limits. Reaching further these limits through technology is not easy. A relatively easy way to drastically increase the capacity is to use smaller cells (micro, pico, femto etc.) and not better technology. While there is some prospect that MIMO might increase capacity beyond these, this prospect seems relatively small, especially its benefits decline in a small-cell environment.

Key Technical Observations: ‘Empirical’ Laws Moore’s law Best industry prediction at present suggests that the growth trends will slow around 2010 and may stop altogether around 2016. Use of multiple parallel processors may allow some further improvement, but they are costly, power hungry and difficult to work with. Steady progress but no key breakthrough is expected in areas such as processing power, hard disk, batteries etc.

Key Technical Observations: ‘Empirical’ Laws Edholm’s law Data rates for three communications categories (wired, wireless and nomadic) increase on similar exponential curves, the slower rates trailing the faster ones by a predictable time lag. Key is its prediction that wired and wireless will maintain a near-constant differential in data rate terms, although nomadic and wired seem to gradually converge at around 2030. The law predicts that, in 2010, 3G, Wi-Fi and office LAN will deliver around 1Mbps, 200 Mbps and 5 Gbps, respectively.

Key Technical Observations: ‘Empirical’ Laws Cooper’s law The number of voice calls carried over radio spectrum has doubled every 30 months for the past 107 years, implying that the effectiveness of spectrum utilization in personal communications has improved a million times, i.e., , since 1950. A 15 times by allocating more spectrum, a 5 times by frequency division, a 5 times by enhancing modulation techniques The lion’s share of the improvement, a 2700 times, was the result of effectively confining individual conversations to smaller and smaller areas by spatial division or spectrum reuse Despite being close to the Shannon limit, there is no end in ever increasing wireless capacity if we are prepared to invest in an appropriately dense infrastructure.

Key Technical Observations: ‘Empirical’ Laws Metcalfe’s law The value of a network equals approximately (or ) where is the number of users of the system. Unlike Moore’s or Edholm’s law, Metcalfe’s does not have a time limit to it. It will likely apply to a wide range of new networks in the future as new types of devices and networks are invented.

Technologies Lowering Cost: Backhaul Cells have to be connected back into the infrastructure via backhaul. More costly as cells gets smaller. Backhaul technologies Cabling (copper, coaxial or fiber optic) Fixed wireless Wireless Mesh There are no significant technological changes expected that can lead to reduced backhaul costs and availability. The exception is potential advances of wireless mesh technology but it requires technological innovation to overcome its capacity and delay problems due to multi hopping and self-interference.

Emerging Communications Techniques Disruption-tolerant network (DTN) Software-defined radio (SDR) Cognitive radio (CR) Opportunistic communications Relays Mesh/ad-hoc network Cross-layer control

Disruption-tolerant Networks (DTN) Provide useable and useful communications across networks that are frequently disconnected and/or has no stable end-to-end path due to mobility, density, attack, disaster or environmental conditions. Use store and forward protocol and the concept of bundle. Can provide increases in both availability and capacity. Form an extremely important communication protocol, but it will be a number of years before operational deployment is practical.

Disruption-tolerant Networks (DTN) Human mobility models End-to-end delay Brownian motion Levy walk Random waypoint

Software-defined Radio (SDR) Many future visions of wireless communications involve multi-modal devices connecting to a wide range of different networks or devices modifying their behavior as they discover new types of network. The current approach, incorporating the chipsets from each of the different standards into the device, works well but it is intrinsically inflexible. SDR is for communication devices to be designed like computers with general-purpose processing capabilities and different software for different communications. In the future this flexibility might enable the more efficient use of the spectrum through rapid deployment of the latest radio technologies. Issues with SDR implementation, particularly at terminal Difficulties in implementing broadband antennas Lack of sufficient processing power Insufficient battery power High cost

Software-defined Radio (SDR) In practice, the benefits of SDR appear relatively minor compared to the issues. The current approach of multi-modal devices works well and will likely always be less expensive than a general-purpose SDR radio. Further, since new technologies are generally introduced much less frequently than users replace handsets, there is little need for a handset to download a new standard. Because of this, we do not expect ‘true’ SDRs that can reprogram their radio at handsets during the next two decades. We do, however, expect handsets to be able to download a wide range of new applications. We also expect SDR base stations that can modify their behavior as hey discover new types of network or standard.

Cognitive Radio (CR) Three approaches to spectrum scarcity amelioration Unlicensed bands: e.g., ISM band Underlay: must operate below the ‘FCC Part 15’ noise limit and must use a very broadband carrier (at least 500 MHz), e.g, UWB Overlay: dynamic usage of previously allocated spectrum when non-allocated users can prove that they will not disrupt the incumbent, e.g., CR CR has been defined by ITU as ‘a radio or system that senses, and is aware of, its operational environment and can dynamically and autonomously adjust its radio operating parameters accordingly’. The premise for CR is the observation that Effectively all the spectrum of interest has been allocated, thereby firmly establishing spectral scarcity Most of the spectrum, in most of the places, most of the time is underutilized CR is sometimes described as ‘frequency-agile radio’.

Cognitive Radio (CR) Spectrum utilization for two of the USA’s busiest cities The net spectrum utilization is 17.4% for Chicago and only 13.1% for New York. This suggests considerable opportunity for the deployment of overlay solutions based on CR.

Cognitive Radio (CR) Will CR work? It may not work well. One of the key challenges is to overcome the hidden terminal problem. The problem can be solved by the base station transmitting ‘beacon’, indicating the spectrum band is free. Such an approach requires central management by the owner of the band including a choice as to whether they wish to allow secondary access and if so under what conditions. Is the spectrum needed? There is little need. 3G operators in 2005 were still typically only using 50% of their spectrum allocation. Additional 3G spectrum was promised at 2.5-2.7 GHz and at UHF after analog TV switch-off. Cellular demand may eventually fall as more traffic flows to W-LANs.

Opportunistic Communications Opportunistic scheduling Exploit multi-user diversity in time and frequency. In a large system with users fading independently, there is likely to be a user with a very good channel at some carrier frequency for each time. Long-term total throughput can be maximized by always serving the user with the strongest channel gain for each frequency. Challenge is to share the benefit among the users in a fair way. User 1 frequency Channel gain Fading channel frequency Channel gain User M

Opportunistic Communications Opportunistic routing Source broadcasts each packet without intended receiver. Learn the set of nodes which actually received the packet. A receiver in the set that is closest to the destination is selected to forward the packet. This continues until the destination receives the packet. Opportunistic routing provides more throughput than conventional routing Each transmission has more independent chances of being received and forwarded. Take advantage of transmissions that reach unexpectedly far. TX counts: In opportunistic routing, (1-(1-0.25)^4)^-1+1 = 2.46 In conventional routing, 4+1 = 5

Relays New generation of cellular requires dense BS deployment for the following reasons. Higher data rates can be attained by a smaller cell and a higher carrier frequency. Transmission at high carrier frequency (> 2GHz) is vulnerable to non-line-of-sight environment such as metropolitan area. However, it is unacceptable due to its high deployment and maintenance cost. A cost-effective alternative is multi-hop relaying approach. Dense deployment of cheap relay stations (RS) with low transmit power Multi-hop wireless connection to BS, forming wireless mesh

Relays Benefits of relays Unknowns and challenges Low cost compared to BS deployment Coverage and fairness enhancement Unknowns and challenges If RS-RS and RS-BS transmissions use the same radio with MS-BS and MS-RS transmissions, total system throughput may decrease. RSs act as additional interference sources to neighboring cells so that ICI becomes more severe and total system throughput may decrease unless ICI is tightly managed. Cross-layer control of wireless mesh network is a big challenge.

Wireless Mesh Networks (WMN) A wireless inter-network of various sub-networks including Wi-Fi networks, cellular networks, WiMax networks, sensor networks etc. A wireless backhaul network for Wi-Fi networks, cellular networks, WiMax networks etc. Many other application areas including community networking, enterprise networking, home networking etc.

Wireless Mesh Networks (WMN) The current 802.11-based mesh technology cannot meet the promise. Insufficient capacity even with multiple channels Unfairness depending on path length No proven wireless multi-hop protocol stack TCP from wired Internet, routing protocols (AODV, OLSR, DSR etc.) from MANET and MAC from Wi-Fi network A clean slate protocol stack that can squeeze most performance out is necessary to meet the promise. Its design involves Understanding of optimal interaction between transport, routing, MAC (link scheduling, power control) and PHY layers Finding distributed algorithms and protocols that can most closely approximate the optimality Understanding of multi-link interference and finding maximal independent link sets in a distributed manner Understanding of optimal interaction between mesh links and access links if they share the same radio

Cross-layer Control Wired multi-hop networks Network utility maximization Link capacity is given and constant Flow control problem at transport layer

Cross-layer Control Lagrangian function Dual problem Dual decomposition Flow control at source Congestion price at link TCP is an approximation of this dual decomposition

Cross-layer Control Wireless multi-hop networks Long-term network utility maximization Link capacity is time-varying and a function of resource control Joint rate, power allocation and link scheduling

Cross-layer Control Lagrangian function Dual problem Dual decomposition Flow control at source Scheduling/power control at link Congestion price at link Joint MAC and transport problem Distributed scheduling/power control is a challenge