Physical Layer Research Trends for 5G Brett T. Walkenhorst, Ph.D. Principal Research Engineer, Georgia Tech Research Institute Adjunct Professor, Georgia Tech, School of Electrical and Computer Engineering brett@gatech.edu 404-407-6525 ATIS 5G Symposium Chicago, IL June 9, 2015 GTRI_B- #
Outline A Little Context GTRI Snapshot My Background Commercial Wireless Trends Physical Layer Research Trends Massive Multiuser MIMO Millimeter Wave Cognitive Communications GTRI_B- # 2
GTRI Snapshot Non-profit R&D organization Owned by Georgia Tech Applied research arm of GT Founded 1934 About 2000 total staff ~1000 research faculty ~350 students Locations in 10 states and DC FY2014 contract awards: $363M GT Colleges & Schools R&D: $325M 3
Millions GTRI Awards History 400 350 300 250 200 150 100 50 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 YTD 4
A Little of What We Do Research Applications: Radar, Electronic Warfare, Communications, Networking, Command and Control, Cyber Security, Robotics, Quantum Computing, etc Technical areas: Systems Engineering, Theoretical Physics, Electromagnetics, Signal Processing, Data Analytics, etc Teaching Professional education courses Some academic courses 5
My Background Professional History BS/MS in EE from BYU RF/DSP work at Lucent, Bell Labs PhD in EE Georgia Tech Research Engineer at GTRI Adjunct Professor at Georgia Tech Relevant Experience in Technical Domains GSM base station hardware design, layout, and testing Physical layer waveform design, analysis, and exploitation Communications systems (hardware/software) design and assessment of system tradeoffs 6
Trends 1G: Basic voice service, analog waveforms 2G: Improved coverage, digital waveforms (GSM, CDMA) 3G: Voice and data 4G: DATA and maybe a little voice, IP protocols (WiMax, LTE) Generation 1G 2G 3G 4G 5G? Data Rates N/A <0.5Mbps 63+Mbps 300Mbps Source: https://www.qualcomm.com/media/documents /files/the-evolution-of- mobile-technologies- 1g-to-2g-to-3g-to-4glte.pdf 7
Where are we headed? Some thoughts on direction Wicked fast (data rate is one of the more visible trends) Super efficient mobile network Converged fiber-wireless Massive Multiuser MIMO Millimeter-wave Everything connected (IoT) Cooperative Communications Cognitive Operation/Cooperation Interference management (e.g. multi-user detection) 8
Some Possible Numbers Wicked fast: 10Gbps Super efficient network: Supports up to 100 billion devices (currently around 5 billion); hundreds to millions of subscribers per cell, many of which are always on Deployment Timeframe: 2020 2030 If data rate is critical (and it is), bandwidth is key and maybe number of antennas http://www.huawei.com/5gwhitepaper/ 9
MIMO Multiple Input Multiple Output Concept has been around since 1998 Deployed in Wifi chips Part of 4G standards Began as data rate improvement (spatial multiplexing) Later extended to diversity MIMO Receiver enhancement (e.g. space-time block coding) and beamforming y = H x +n RX Antennas Wireless Channel Channel Matrix (H) TX Antennas MIMO System Diagram 4x4 Example MIMO Transmitter 10
Massive Multiuser MIMO 100 or more antennas on a base station Spatial multiple access (and improved SNR and throughput) via complex beamforming Spatial multiplexing with remaining degrees of freedom (up to what the channel will support) Massive MIMO antenna for 5G communications research Image from http://forums.xilinx.com/t5/xcell-daily-blog/softwaredefined-radio-dev-platform-for-5g-research-handles-mimo/bap/441754 11
Conceptual Views of Massive MIMO One conceptual view of massive MIMO Image from: http://nutaq.com/en/blog/massive-mimotechnology-big-shift-next-generation-wirelessbroadband-communications Another view Image from: http://blog.3g4g.co.uk/search/label/mimo 12
Millimeter-wave Seeking to drastically improve data rates, spectrum is the biggest crunch Options being discussed Confluence of fiber and wireless Smaller cell sizes with greater frequency reuse in dense urban environments Millimeter-wave Lots of options for carrier frequency Large instantaneous bandwidth Image from: J. Laskar, et al, The next wireless wave is a millimeter-wave, Microwave Journal, Aug 2007. 13
Propagation Challenges of mmwave Free space spreading loss is a function of wavelength For fixed antenna gains, we get very large spreading losses Omnidirectional links Short range For fixed antennas sizes, we get very directional antennas We can close our links better, but we need a way to handshake and then closely track CSI so we can point our beams Shadowing: physical structures start to look opaque at these wavelengths This can work with very small cells Indoor augmentation of terrestrial outdoor network 14
Massive MIMO and mmwave Combination of massive MIMO and millimeterwave Presents interesting challenges similar to search/track radar If we insist on maintaining reasonable outdoor link ranges Higher atmospheric attenuation leads to smaller range Ok in some environments Especially if networks move toward smaller cells 15
Cognitive/Adaptive/Cooperative/[insert favorite buzzword] Communications Cognitive radio term coined in 1998 Leveraging advances in the flexibility offered by software defined radio, Joe Mitola suggested radios could act more intelligently Radios can work together to understand their environment and act accordingly Potential for more efficient use of spectrum PCAST report to President, July 2012, suggests methods for pushing this concept toward commercialization City-wide testbeds Database subscription These are good first steps, but a database approach offers extremely slow response times Potential exists for much faster execution 16
Cooperative Spectrum Sensing Using radios mounted on robots, we collect data and interpolate Goal is to map EM emissions in space and frequency Y Position X Position 17
Channel Gain Mapping Actual Channel Gain Map Example: Kriged Kalman Filtering Where will my signal, my buddy s signal, and the PU s signal propagate? Estimated Channel Gain Map Images from: S. Kim, E. Dall Anese, G.B. Giannakis, Cooperative spectrum sharing for cognitive radios using kriged kalman filtering, IEEE Journal of Selected Topics in Signal Processing, vol. 5, no. 1, pp. 24-36, Feb 2011. Equations from: D. Lee, S-J. Kim, Channel gain cartography via low rank and sparsity, IEEE Asilomar Conference, pp. 1479-1483, Nov 2014. 18
Tomography and Applications Notional EM Tomography scheme Image from: J Wilson, N Patwari, Radio tomographic imaging with wireless networks, IEEE Transactions on 9 (5), 621-632 η Bg w [Wilson et. al 08] Field position Attenuation SLF @ position s Spatial weight function Tx position Rx position 19
Tomography Results 20
Why? = Primary User = Cognitive Radio Concept: Mobile radio (or radio relay) moves to avoid interference and increase link quality. Triangle moves to lowinterference/high-throughput position 21
Cognitive Summary All of the above algorithms utilize simple data sharing among nodes RSSI Node position estimate They solve distributed optimization problems to yield useful information EM Physical Many questions remain as to how such a network of cooperative nodes would be employed 22
Summary Massive MIMO Spatial multiplexing boosts data rates Beamforming allows for multiuser support and makes mmwave more feasible Millimeter wave Spectrum is the biggest problem This opens a large swath of spectrum for us to use if we can tackle the technical challenges Cognitive communications Adapt for more efficient use of spectrum Machine learning algorithms can optimize parameters on timescales that no human can handle 23
Questions? Contact Info: Brett T. Walkenhorst, Ph.D. brett@gatech.edu 404-407-6525 24