Cognitive Radio Networks Narayan B Mandayam WINLAB Rutgers

Cognitive Radio Networks Narayan B. Mandayam WINLAB Rutgers University NSF Workshop on Bridging the Gap between Wireless Networking Technologies and Advances at The PHY Layer August 27, 2007 1

Cognitive Radio Research A Multidimensional Activity n Spectrum Policy Economics ¨ Regulation ¨ Legal ¨ Business ¨ n n Theory and Algorithms Cooperative Communications ¨ Information & Coding Theory ¨ Statistical Signal Processing ¨ Game Theory & Microeconomics ¨ Hardware/Software Platforms & Prototyping Programmable agile radios ¨ GNU platforms ¨ Cognitive Radio Network Testbeds ¨ 2

The Spectrum Debate Triumph of Technology vs. Triumph of Economics n Open Access (Commons) ¨ [Noam, Benkler, Shepard, Reed …] n n Spectrum Property Rights ¨ [Coase, Hazlett, Faulhaber+Farber] n n n Triumph of Technology Agile wideband radios will dynamically share a commons Success of 802. 11 vs 3 G Triumph of Economics Owners can buy/sell/trade spectrum Flexible use, flexible technology, flexible divisibility, transferability A spectrum market will (by the force of economics) yield an efficient solution What everyone agreed on (~ 10 years ago): n n Spectrum use is inefficient FCC licensing has yielded false scarcity 3

The Spectrum Debate & Cognitive Radio n What everyone agrees on now: Spectrum use is inefficient ¨ FCC licensing has yielded false scarcity ¨ n Possible middle ground? Dynamic spectrum access ¨ Short-term property rights ¨ Spectrum use driven by both technology and market forces ¨ n Cognitive Radios with ability to incorporate market forces? ¨ Microeconomics based approaches to spectrum sharing n “Dynamic Spectrum Access Models: Towards an Engineering Perspective in the Spectrum Debate” by Ileri & Mandayam, To appear in IEEE Communications Magazine 2007 4

Motivation for Dynamic Spectrum and Cognitive Radio Techniques: n Static allocation of spectrum is inefficient ¨ n Spectrum allocation rules that encourage innovation & efficiency ¨ n Unlicensed systems need to scale and manage user “Qo. S” Density of wireless devices will continue to increase ¨ n Free markets for spectrum, more unlicensed bands, new services, etc. Anecdotal evidence of WLAN spectrum congestion ¨ n Slow, expensive process that cannot keep up with technology ~10 x with home gadgets, ~100 x with sensors/pervasive computing Interoperability between proliferating radio standards ¨ Programmable radios that can form cooperating networks across multiple PHY’s 5

Towards Cognitive Radio Networks Research themes that have emerged from mobile ad hoc and/or sensor networks research: n Hierarchical Network Architecture wins ¨ n Cooperation wins ¨ n Achievable rates via information theoretic relay and broadcast channels “Global” awareness and coordination wins ¨ n Capacity scaling, energy efficiency, increases lifetimes, facilitates discovery Space, time and frequency awareness and coordination beyond local measurements Efficient operation requires radios that can: Cooperate ¨ Collaborate ¨ Discover ¨ Self-Organize into hierarchical networks ¨ 6

Cognitive Radio - Theory & Algorithms Fundamental research and algorithms – based on foundations of: ¨ Information and Coding Theory n ¨ Signal Processing n ¨ Collaborative signal processing, Signal design for spectrum sharing, Interference avoidance, Distributed sensing algorithms Game Theory n ¨ Relay cooperation, User Cooperation, Coding techniques for cooperation, Collaborative MIMO techniques Spectrum warfare, Microeconomics and pricing based schemes for spectrum sharing, negotiation and coexistence, Coalition formation, Incentive mechanisms for cooperation MAC and Networking Algorithms n Discovery protocols, Etiquette protocols, Self-organization protocols, Multihop routing 7

Cognitive Radio: Reactive Schemes n Reactive (autonomous) methods used to avoid interference via: Frequency agility: dynamic channel allocation by scanning ¨ Power control: power control by interference detection and scanning ¨ Time scheduling: MAC packet re-scheduling based on observed activity ¨ Waveform agility: dynamism in signal space ¨ n Reactive schemes (without explicit coordination protocols) have limitations: Interference is a receiver property! C B D’s agile radio waveform without coordination protocol D Coverage A’s agile radio waveform A cannot hear D area of D A Y with coordination Coverage area of A Hidden Terminal Problem 8

Cognitive Approaches: Outlook n Cognitive radio networks require a large of amount of network (and channel) state information to enable efficient Discovery ¨ Self-organization ¨ Cooperation Techniques ¨ End-to-end routed path From A to F Bootstrapped PHY & control link C PHY A B B DD PHY C E PHY B A A Control (e. g. CSCC) F Multi-mode radio PHY Ad-Hoc Discovery & Routing Capability Functionality can be quite challenging! 9

Cognitive Radio: Design Space n Broad range of technology & related policy options for spectrum Unlicensed band + simple coord protocols Protocol Complexity (degree of coordination) Ad-hoc, Multi-hop Collaboration Internet Server-based Spectrum Etiquette Unlicensed Band with DCA (e. g. 802. 11 x) Internet Spectrum Leasing Static Assignment “cognitive radio” schemes Radio-level Spectrum Etiquette Protocol Reactive Rate/Power Control Agile Wideband Radios “Open Access” + smart radios UWB, Spread Spectrum Hardware Complexity 10

Cognitive Radios need help too! n n Infrastructure that can facilitate cognitive radio networks Examples of coordination mechanisms: ¨ Information aids n ¨ “Spectrum Coordination Channel” to enable spectrum sharing Network architectures n “Spectrum Servers” to advise/mediate sharing 11

Cognitive Radio: Spectrum Policy Server n Internet-based Spectrum Policy Server can help to coordinate wireless networks (a “Google for spectrum”) Needs connection to Internet even under congested conditions (. . . low bit-rate OK) ¨ Some level of position determination needed (. . coarse location OK? ) ¨ Spectrum coordination achieved via etiquette protocol centralized at server ¨ Spectrum Policy Server www. spectrum. net Internet AP 1 Access Point (AP 2) WLAN operator A Etiquette Protocol AP 1: type, loc, freq, pwr AP 2: type, loc, freq, pwr BT MN: type, loc, freq, pwr WLAN operator B Master Node Wide-area Cellular data service Ad-hoc Bluetooth Piconet 12

What can a Spectrum Policy Server do? rate 1 rate 2 n rate 4 Spectrum Server rate 3 Spectrum Server facilitates co-existence of heterogeneous set of radios by advising them on several possible issues: q. Spectrum policy q. Interference information q. Scheduling and coexistence q. Location specific services q. Context q. Mobility Management q. Addressing q. Authentication q. Security q. Content 13

Cross-layer scheduling of end-to-end flows (( )) AP Flow 1 Flow 2 Wireless network Overarching design principles for wireless networks • Physical links - achievable rates depend on the PHY layer transmission employing a variety ofatphysical schemes, signal processing receiver etc. layer strategies? • MAC – distributed/centralized schemes to avoid/control interference • Routing – decision made based on metric specified by application running on the network 14

Scheduling with a Spectrum Server (Raman-Yates-Mandayam) 1 4 1 Glk = link gain from Tx k to Rx l 2 3 3 network with 4 links Transmission mode [1 0 1 0] Rate matrix C = [6. 6 [0 [0 [0 0 6. 6 0 0 0. 01 0. 06 0 0 6. 6 0 0. 56 0 1. 0 0 0 1. 86 0 0. 01 0 0. 06 0 0. 83 0 0 6. 65 § Spectrum server specifies § xi = fraction of time mode i is ON §Average rate in link l is rl = i cli xi 2. 05 0 0 0. 32 0 0. 97 0 0. 05 0. 01 0 0. 06 0 0 0. 04 0. 40 0. 49 0 0. 01] 0 0. 77 0. 06 ] 0. 04 ] 0. 19 0. 05 0. 04] The rate matrix C includes the achievable physical layer rates for a wide rate of physical layer transmission techniques 15

Universal cross-layer scheduling framework – centralized approach Utilities: Max throughput, max-min fairness, proportional fairness, energy efficiency Maximize User utilities such that: Physical layer rates c 11 . . . c. M 1 . cli . x. M c. L 1 PHY rates x 1. x 3 x 2 MAC schedule x 1 > 1 0 1 1 1 … 0 f 1. . … 0 NETWORK routes . f. M Higher layer flow requirements TRANSPORT flows f 1 x 4 f 2 16

Cognitive Radio Network Experiments Hardware/Software Platforms@WINLAB n n ORBIT radio grid testbed currently supports ~10/USRP GNU radios, 100 low-cost spectrum sensors, WARP platforms, WINLAB Cognitive platforms and GNU/USRP 2 Cross-layer design and experimentation Suburban ORBIT Radio Grid Current ORBIT sandbox with GNU radio 20 meters 500 meters Office 30 meters Urban 300 meters Radio Mapping Concept for ORBIT Emulator 400 -node Radio Grid Facility at WINLAB Tech Center Programmable ORBIT radio node Planned upgrade (2007 -08) URSP 2 CR board 17

Wireless and the Future Internet n “Wireless” is overtaking “Wired” as the primary mode of connectivity to the internet ~750 M servers/PC’s, >1 B laptops, PDA’s, cell phones, sensors ~500 M server/PC’s, ~100 M laptops/PDA’s Wireless Edge Network INTERNET Wireless Edge Network 2005 2010 §Wireless usage scenarios that will impact future internet design q. Mobile data applications q. Multihop Mesh networks q. Sensor and vehicular networks 18

Wireless/Mobile/Sensor Scenarios and the Future Internet n Some architectural and protocol implications for the future Internet. . . ¨ Integrated support for dynamic end-user mobility ¨ Wireless/mobile devices as routers (mesh networks, etc. ) Network topology changes more rapidly than in today’s wired Internet Significant increase in network scale (10 B sensors in 2020!) New ad hoc network service concepts: sensors, P 2 P, P 2 M, M 2 M, … Addressing architecture issues – name vs. routable address Integrating geographic location into routing/addressing Integrating cross-layer and cognitive radio protocol stacks Data/content driven networking for sensors and mobile data Pervasive network functionality vs. broadband streaming Power efficiency considerations and computing constraints for sensors Many new security considerations for wireless/mobile Economic incentives, e. g. forwarding and network formation ¨ ¨ ¨ 19

NSF GENI Implementation Wireless Sub-Networks Overview Location Service Emerging 5 Technologies (cognitive radio) Advanced Technology Demonstrator (spectrum) Broadband Services, Mobile Computing Other GENI services Infrastructure Ad-Hoc Mesh 2 Network 1 NSF Radio Testbeds Open API 3 Wide-Area Networks “Open” Internet Concepts for Cellular devices Sensor 4 Networks Embedded wireless, Real-world applications Protocol & Scaling Studies Emulation & Simulation 20

Cognitive Radio in NSF’s GENI Project n Propose to build advanced technology demonstrator of cognitive radio networks for reliable wide-area services (over a ~50 Km**2 coverage area) with spectrum sharing, adaptive networking, etc. Basic building block is a cognitive radio platform, to be selected from competing research projects now in progress and/or future proposals ¨ Requires enhanced software interfaces for control of radio PHY, discovery and bootstrapping, adaptive network protocols ……. . suitable for protocol virtualization ¨ FCC experimental license for new cognitive radio band ¨ Cognitive Radio Network Node Cognitive Radio Client Spectrum Server Cognitive Radio Network Node Cognitive Radio Client Spectrum Monitors Connections to GENI Infrastructure Research Focus: 1. New technology validation of cognitive radio 2. Protocols for adaptive PHY radio networks 3. Efficient spectrum sharing methods 4. Interference avoidance and spectrum etiquette 5. Dynamic spectrum measurement 6. Hardware platform performance studies 21