Viral Radio Adaptive and cooperative exploitation of RF photons

Transcription

1 Viral Radio Adaptive and cooperative exploitation of RF photons David P. Reed Adjunct Professor, MIT Media Lab MIT Communications Futures Program

2 Technical basis of viral communications Software Defined, Cognitive Radios allow decentralizing function and adaptation to intelligent edges of the network DSP Generates and Recognizes Waveforms Software Antennas (configurable materials and structures) Download enables upgrade (for a while) Adaptive and distributed control Distributed sensing of propagation environment Distributed control in response to real-time demand System adaptation and evolution costs drop to near-zero Copyright 2002, 2003, 2004 David P. Reed 2

3 A Society of Cognitive Radios Viral network definition: each new user preserves or increases capacity and other economic value to existing users, and benefit to new user increases with scale of existing network Cognitive radios that can cooperate to optimize value under actual demand, while behaving politely to radio systems with more limited capabilities Copyright 2002, 2003, 2004 David P. Reed 3

4 Shannon s framework: bits and C = W log( 1+ channels P N 0W C = capacity, bits/sec. W = bandwidth, Hz. P = power, watts N 0 = noise power, watts/hz. ), due to ClaudeShannon Capacity Channel capacity is roughly proportional to bandwidth, and logarithm of power. Bandwidth

5 There is no cat-5 RF is not a virtual wire Copyright 2002, 2003, 2004 David P. Reed 5

6 Exploiting variability Copyright 2002, 2003, 2004 David P. Reed 6

7 Propagation cannot be planned Copyright 2002, 2003, 2004 David P. Reed 7

8 Transport Capacity: One important measure of radio network utility Network of N stations (transmit & receive) Scattered in a fixed space Each station chooses randomly to send messages to other stations What is achievable total transport capacity, C T, in bit- meters/second? b d s, r s, r = bits from s to r = distance C b s r N T =, from s, r t s d s, r to r

9 One example of an architectural improvement: hop-by by-hop repeating Energy/bit reduced by 1/hops. Many paths can operate concurrently. What is repeater network s capacity as radios are added?

10 Achievable operating region: A measure of architectural A->B capability A->C N 2 dimensions (end-to to-end pairs) Quality measure: performance power efficiency end-to to-end latency C->B Copyright 2002, 2003, 2004 David P. Reed 10

11 A Viral Architecture Framework Each radio extends orthogonal propagation space Each radio adds computational capacity at edge Cooperation allows collective capacity to be dynamically allocated, benefiting all in available capacity to individuals Cooperation gain Disperse communications load widely, and if possible, fairly Cooperative evolution from legacy systems Copyright 2002, 2003, 2004 David P. Reed 11

12 End-to to-end argument The endpoints have the best information about demand, so are in the best place to manage the resource Minimal function in the network to manage the resource: magic mirror, distributed load sensing, edge-based dynamic allocation Copyright 2002, 2003, 2004 David P. Reed 12

13 Elements and decisions Copyright 2002, 2003, 2004 David P. Reed 13

14 Decentralized Control of Propagation Space Flux routing not yes/no interference but tradeoffs in achievable rate region Distributed sensing of propagation Local decisions, incorporating global information Decide while communicating about actual conflict vs. potential conflict (distributed sensing of demand) Copyright 2002, 2003, 2004 David P. Reed 14

15 Inputs to local decisions Traffic in out through (movable) Neighborhood traffic capability propagation reach Energy availability Radio capabilities (heterogeneity) Copyright 2002, 2003, 2004 David P. Reed 15

16 Analogy: Edge-based Congestion Control in IP Internet edge-based congestion control Congestion = queue size Queue grows drop packets Drop packet detected at edge, which reduces rate Collective action manages congestion Counterintuitively drop first vs. drop last Early drop (simulate drop) May prioritize drops by QoS latency or loss rate) Edge-based flux routing control Congestion = local incident energy Energy grows stop reflecting Signal quality loss detected at edge,enlisting others Collective action spreads load in space and time Copyright 2002, 2003, 2004 David P. Reed 16