Cooperative Relaying Networks

Size: px

Start display at page:

Download "Cooperative Relaying Networks"

Diana Warren
5 years ago
Views:

1 Cooperative Relaying Networks A. Wittneben Communication Technology Laboratory Wireless Communication Group Outline Pervasive Wireless Access Fundamental Performance Limits Cooperative Signaling Schemes Joint Cooperative Diversity and Scheduling The Rich Array/Poor Scattering Regime The RACooN Laboratory

2 Pervasive Wireless Access Networks cellular: GSM UMTS WPAN WLAN RFID Internet backhaul Pervasive wireless access Sensor network Bluetooth WMAN Heterogeneous nodes RFID tags, readers sensors, actors communication appliances information access information processing backhaul access points... Heterogeneous standards IEEE WLAN IEEE WPAN IEEE WMAN (Hiperlan) Bluetooth DECT various RFID.. Available spectrum (approx.) (ISM) (ISM) (ISM) (ISM) Some Wireless Access Systems 1000M link 100M throughput [bps] 10M 1M 100k 10k Body Area Networks WPAN 15.3a 15.3 Sensor Networks 11a next generation WLAN g 11b 15.1Bluetooth ZigBee WLAN spatial multiplexing f 0 beyond 5 GHz 1k RFID range [m]

3 Hierarchical Heterogeneous Nodes information access, peripherals sensors, actors tags, sensors internet (backhaul) access information processing Network characteristics hierarchical nodes node density spot coverage uncoordinated, unlicensed ad hoc infrastructure Design objectives data rate, QoS range position location low cost low EM exposure Existing systems are insufficient Outline Pervasive Wireless Access Fundamental Performance Limits Cooperative Signaling Schemes Joint Cooperative Diversity and Scheduling The Rich Array/Poor Scattering Regime The RACooN Laboratory

4 Capacity of Wireless Networks (Gupta/Kumar, Trans. On IT, 000) n nodes optimally placed Each node can transmit at W bits/sec Traffic patterns, ranges, powers are optimally assignes Point-to-point coding Gaussian interference model (no joint decoding) Main Result: Order of the aggregate throughput capacity is ( ) W λ( n) =Θ n = Θ n W bit/sec n Capacity of Wireless Relay Networks (Gastpar/Vetterli, Infocom 00) n Nodes randomly distributed over a disk Source and destination randomly chosen Ever node can hear every other node Source transmits only half the time Relay traffic pattern with one active source-destination pair Gaussian channels Arbitrary complex network coding is used source Main result: C = log n destination average per-node power constraint coherent combining on downlink

5 Maximizing Degrees of Freedom in Wireless Networks (Borade et al., Allerton 003) Two-hop network Broadcast from multi-antenna source (beamforming) Multiple access to multi-antenna destination (V-BLAST) multihop with MIMO intermediate nodes Degrees of Freedom in Wireless Networks (Borade et al., Allerton 003) Multi-hop network Amplify-and-forward relays Establishes a distributed point-topoint MIMO channel Source uses same codebook as for a MIMO system (Gaussian codebooks) For fixed k and n the system achieves for high SNR a rate R(SNR) nlog(snr) P SNR = σ

6 Degrees of Freedom in Wireless Networks (Borade et al., Allerton 003) No communication is possible when k (number of hops) and SNR is fixed Full degrees of freedom are achieved when k is fixed and SNR Question: For which functions k n (SNR) full n degrees of freedom can be achieved? Answer: k (SNR) lim n = 0 log(snr) SNR k e SNR 4.3 db / hop Outline Pervasive Wireless Access Fundamental Performance Limits Cooperative Signaling Schemes Joint Cooperative Diversity and Scheduling The Rich Array/Poor Scattering Regime The RACooN Laboratory

7 Cooperative Diversity Data Link Physical LLC MAC diskret analog decode, forward filter, amplify, forward amplify, forward basestation user 1 user Distributed Antenna Uplink Scenario R1 user 1 user R central processor P 1 P * k 10 * k 0 Z 0 + Y 0 R 1 achievable rate region basestation P k k H k Z 0 X X X X + R perfect CSI at transmitter beamforming (coherent combining) Y 0 P k R= R1 + R < C σ Z P = P1 + P

8 Multi-Access Uplink Scenario achievable rate region for P = P = P 1 / W ; R user X 1 P 1 * k 10 X Z 0 R 1 + user W ; R X X * k 0 Y 0 P basestation R coherent combining not possible due to independent codebooks with CSI: power loading P k R= R1 + R < C σ Z P k 10 R1 < C σz P k 0 R < C σz User Cooperation Diversity W 10 X10 P 10 W + 1 PU 1 k10/ k10 W 1 U X X * k 10 [Sendonaris et al 98/0] achievable rate region for P = P = P 1 / W Z 0 R 1 W W W 1 0 PU k0/ k0 U X X0 P 0 X + * k 0 + Y 0 basestation W = W1 + W W = 0 R perfect CSI users share a part of their data this part is transmitted coherently with the same codebook

9 Outline Pervasive Wireless Access Fundamental Performance Limits Cooperative Signaling Schemes Joint Cooperative Diversity and Scheduling The Rich Array/Poor Scattering Regime The RACooN Laboratory Multiuser diversity vs. low mobility In a large wireless network the probability is high that the base station can serve one high-data rate user -> multiuser diversity aggregate throughput (system throughput) can be maximized by always serving the user with the strongest channel disadvantage: in low mobility environments channel variations are not sufficiently large enough -> high delays at some user nodes

10 Fairness and delay 35 SNR of asymmetric user channels SNR [db] User 1 User time slots in asymmetric channels (near-far situation) adaptive scheduling is unfair => high delays even in high mobility environment challenge is to achieve multiuser diversity gains while providing certain amount of fairness Proportional Fairness Question: How to achieve fairness among users with different fading statistics? Solution: Serve user who has best SNR compared to its average SNR (within a given latency time-scale t c ) Comments: proportional fair scheduler normalizes the SNR of each user to a similar average value scheduler operates away from aggregate throughput optimum

11 Proportional Fairness and Low Mobility Environments 4 SNR of asymmetric user channels 3.5 normalized SNR [db] note different y-scale User 1 User time slots in low mobility environments channel variations are not sufficient to achieve multiuser diversity gains with fairness introduce channel fluctuations artificially? Opportunistic Beamforming using dumb antennas (Viswanath, Tse, and Laroia, Trans. Inf. Theory 00) basestation with multiple antennas time-variant weights at each antenna SNR feedback from all users randomly swept beam and opportunistically send data to best user

12 Slow Fading Environment: Before and after Before After artificially introduced high mobility (time-variance) Performance for large number of users performance of true beamforming is achieved less feedback and channel measurements required

Distributed Relay networks amplify-and-forward relays introduce channel fluctuations by time-variant amplification gain at relays Joint Cooperative Diversity and

13 Distributed Relay networks amplify-and-forward relays introduce channel fluctuations by time-variant amplification gain at relays Joint Cooperative Diversity and Scheduling (Wittneben/Hammerström Globecom 004) 1% aggregate outage throughput is improved by a factor of nine if six active source/destination pairs are considered

14 Outline Pervasive Wireless Access Fundamental Performance Limits Cooperative Signaling Schemes Joint Cooperative Diversity and Scheduling The Rich Array/Poor Scattering Regime The RACooN Laboratory Array and Propagation Model Number of antenna elements ( ) ( ) N N f /f N f with N = 16 A/ π λ a a,0 0 a,0 N a,0 0 A λ / Power path loss ( ( )) ( ) x = λ/ 4π d G G x d/d γ PL 0 TX RX PL,0 0 d d 0 ( ( )) x = λ/ 4π d G G x b /f PL 0 TX RX PL,0 N in the sequel b=1 without loss of generality fixed distance

15 Some Capacity Considerations Channel matrix: HSD = HSD,N b/fn H SD,N /fn ( ) singular values{ σ k } SD N a k ( ) Instantaneous capacity: CSD ( HSD ) = log 1+ ( P s /N a ) ( σsd / σw ) no power loading per complex dimension Ergodic capacity: C E C ( H ) = SD HSD SD SD k= 1 ( ) () () No scattering: σ 1 1 CSD = log ( 1+ Ps N a,0 / σ SD,N = Na σ SD = N a /fn w ) HSD,N [ m,n] = 1 Rich scattering: HSD,N [ m,n] = CN( 0,1) p( σ N ) ( ) = ( + σ ( ) ( σ σ )) CSD Na fn Na,0 E log 1 P s /f N N / w N a N : one-dimensional pdf of singular values of HSD,N 1/ Na essentially independent of Na Asymptotic value f ;N : power limited regime N a for Na 4 ( ) ( ) ( ) C = N /ln P / σ A P / σ SD a,0 s w s w Rich array/poor scattering paradigm capacity [bit/channel use] no scattering Cf0_v1.m reality rich scattering? would require > 56 relevant scatterers poor rich scattering poor rich array normalized carrier frequency f N

16 Active Scatterer Concept (Wittneben/Rankov, IST 003, SPAWC 004) Example: Wireless Distribution System for WLAN Network operates in infrastructure mode Channels are domiated by line-ofsight (WLAN at 4 GHz) Idle nodes as relays (active scatterers) to introduce artifical multipath structure into effective channel Active Scatterer Concept (Wittneben/Rankov, IST 003, SPAWC 004) One source/destination pair with N antennas each K amplify-and-forward relays Channel shaping via active scattering for capacity gains Linear increase in capacity when N<K spatial multiplexing gain Logarithmic increase in capacity when N>K array gain

17 Distributed Antenna System (DAS) Scenarios (Wittneben/Rankov, URSI-EMTS 004, VTC Fall 004) Access point: Na = 4 fn equispaced support nodes 1.5m destination Propagation model: γ = no scattering source Source, destination: random uniform placement stationary no antenna coupling central processor DAS: -hop traffic pattern decode DAS: decode&forward linear DAS: linear processing; simple link adaptation Performance of LDAS and DDAS versus the carrier frequency f N case6 Nr= 64 NaS= 64 NaD= 64 krice= r:t1 b:ta g:tb m:tc k:t % outage capacity DDAS LDAS total power constraint at source and support nodes no power loading across spatial subchannels further details in [WittRank04] pp normalized frequency f N DAS efficiently exploits rich array/poor scattering regime [WittRank04]: Distributed Antenna Systems and Linear Relaying for Gigabit MIMO Wireless, VTC Fall 004

18 Antenna spacing (16x16x16)-system 10% outage capacity case Nr= 16 NaS= 16 NaD= 16 krice= r:t1 b:ta g:tb m:tc k:t0 45 DDAS 40 LDAS pp 15 f N = const robust performance residual spatial multiplexing gain for pp normalized antenna spacing d a /λ compact antenna arrays feasible Number of source/destination antenna elements (N a x16xn a )-system 10% outage capacity case3 Nr= 16 NaS= 16 NaD= 16 krice= r:t1 b:ta g:tb m:tc k:t DDAS 35 LDAS pp 10 5 f N = const LDAS: downlink eigenbeams known DDAS: no downlink CSI rich tradeoff residual spatial multiplexing gain for pp #source antennas N a excellent performance of LDAS

Scheduling The Rich Array/Poor Scattering Regime The RACooN

19 Outline Pervasive Wireless Access Fundamental Performance Limits Cooperative Signaling Schemes Joint Cooperative Diversity and Scheduling The Rich Array/Poor Scattering Regime The RACooN Laboratory RACooN Laboratory at ETH Zurich 10 nodes 5-6GHz 80MHz bandwidth

20 Initial Measurements Relay magnitude of impulse response Direct x 10-8

Opportunistic Communication in Wireless Networks

Opportunistic Communication in Wireless Networks David Tse Department of EECS, U.C. Berkeley October 10, 2001 Networking, Communications and DSP Seminar Communication over Wireless Channels Fundamental