Sei sulla pagina 1di 55

Multi-hop and MIMO Relay

Robert W. Heath Jr.

Wireless Networking and Communications Group (WNCG)


Dept. of Electrical and Computer Engineering
The University of Texas at Austin
Austin, TX, USA
2007 Huawei May 2007

Multi-Hop Network Architectures

Coverage an issue in high-capacity


wireless systems particularly in
Mobile urban jungles and cluttered terrain
Station - Signal degradation due to distance
- Obstructions interfere with signals

Base Station
Multi-hop network architecture used
Fixed Relay to extend signal reach and bypass
obstructions
Mobile Station /
Mobile Relay Fixed and mobile relays provide
Mobile alternate paths between user
Station
Direct path device and base station
Alternate path

2007 Robert W. Heath Jr. 2


2007 Huawei May 2007

Throughput Ehancement w/ Relay


NLOS, Fixed Relay LOS, Fixed Relay

(Source: [ComMag, Sept. 2004])

Relay enables enhanced system throughput in both NLOS and


LOS environments

2007 Robert W. Heath Jr. 3


2007 Huawei May 2007

Types of Relays & Coordination


RF repeater
- Basic RF amplify and forward device
- No coordination requirements, doesnt work well with MIMO
Fixed relay
- Fixed infrastructure, not connected to network like BS
- Always available, used to fill in dead spots
- May require moderate coordination between BS and MS
Mobile relay
- Other MS acts as a relay
- Only works when there are many subscribers
- Requires substantial coordination between MS stations
MIMO can be used in different configurations

2007 Robert W. Heath Jr. 4


2007 Huawei May 2007

Functional Comparison
Mobile Relay Fixed Relay RF Repeater

Coverage Extension
O O O
Overcoming Dead-spot

Power Consumption
O
Reduction

Capacity Increase O

Selective Relay O O X

Amplifying O O O

Physical Layer O ? X

MAC Layer O ? X

Routing O X X
Mobility Support (needed) (not needed) (not needed)

2007 Robert W. Heath Jr. 5


2007 Huawei May 2007

Fixed Relay vs. Mobile Relay


Fixed relays
- Practical protocols that include relaying
- Amplify and forward versus decode and forward
- Handoff, relay selection needs work
- ARQ and retransmission at RS? At BS?

Mobile relays
- Practical protocols are much more complex
- No service guarantee
- Possibly increased energy consumption (lower battery life)
- Additional hardware and functionality (higher terminal cost)
- Security issues
- Handoff and relay selection more complex
- Need incentives for users: symmetric cooperation? Air time offers?

Expectations in 4G networks
- Fixed relays will be developed and deployed soon
- Research will continue on mobile relaying, later deployment

2007 Robert W. Heath Jr. 6


2007 Huawei May 2007

Different Relay Strategies

Network Coding

Virtual Antenna Array

Distributed Mobile
Coverage

Space-time Coding

MIMO MU-MIMO
Relay Relay Collaborative Fixed
MIMO Diversity Spatial
Multiplexing
No relay

MIMO MU MIMO
SISO Multiplexing
Multiplexing

End-to-end sum data rates

Several different tradeoffs between coverage and capacity

2007 Robert W. Heath Jr. 7


2007 Huawei May 2007

Overview of Relay Strategies (1/3)


Distributed space-time coding
- Cooperative diversity
- Nodes collaborate to transmit space-time codeword
Collaborative spatial multiplexing
- Nodes transmit independent symbols in each time
slot
MIMO Relay
- One transmitter, one receiver, and one relay
- All nodes have multiple antennas

2007 Robert W. Heath Jr. 8


2007 Huawei May 2007

Overview of Relay Strategies (2/3)


Multiuser MIMO Relay
- Point-to-point MIMO link between tx and relay
- MIMO broadcast link between relay and rx stations
Virtual antenna arrays (uni-directional)
- Nodes collaborate to send source message
Also supports multi-source transmission
- General concept includes the above 4 strategies

2007 Robert W. Heath Jr. 9


2007 Huawei May 2007

Overview of Relay Strategies (3/3)


Network coding
- Bi-directional, analog
- Two nodes use relay to transmit information to each
other
- Increased spectral efficiency by using intelligent
coding at the relay
- Also known as bi-directional or two-way relaying

2007 Robert W. Heath Jr. 10


2007 Huawei May 2007

Distributed Space-Time Coding


(Cooperative Diversity)
RS/MS

BS collaboration

RS/MS

Space-Time Block Encoder


Mobiles / relay exchange information
Perform space-time coding as if they were collocated
- Alamouti code is good building block
Provides additional diversity gain
Requires a coordination protocol, spectrum, etc.

2007 Robert W. Heath Jr. 11


2007 Huawei May 2007

Cooperative Diversity Protocols


Amplify-and-Forward Decode-and-Forward S
S

R R

D D

Amplifying factor - Full decoding at the relay

Selection Relaying S
Incremental Relaying S

R R

D D
Function of S-R lin Function of S-D lin
k k
- Threshold test at the relay - ARQ in the relay context
- More efficient use of resource

2007 Robert W. Heath Jr. 12


2007 Huawei May 2007

Amply-Forward vs. Decode-Forward @ RS

DF RS
Achievable Rate

d2
<

AF

BS d1 MS

>
! : path loss attenuation coefficient
d1

Achievable rate comparison


- For large d1, AF > DF
- For small d1, DF > AF
Need to investigate sum rate analysis with practical MIMO solutions

2007 Robert W. Heath Jr. 13


2007 Huawei May 2007

Distributed Space-Time-Coded Protocols


D D

1 1
2 2
R R
R m R m
R R

S S

Phase I Phase II

Relay transmission strategies in the second phase


- Repetition on orthogonal subchannels [Laneman et al., IT03]
- Decode-and-forward space-time code (DF-STC) [Laneman et al., IT03]
[Lampe et al., TCOM06 ]
Full CSI at each relay
- Amplify-and-forward space-time code (AF-STC) [Lanemane et al., IT03]
No CSI at each relay
- Opportunistic amplify-and-forward [Bletsas, Ph.D 05]

2007 Robert W. Heath Jr. 14


2007 Huawei May 2007

Key Results on DSTC (1/3)


Decode and forward Amplify and forward

Depends on number of
Diversity Gain relays which decode Full: min(nt, nr)Xm
correctly => min(nt, nr)Xm

Moderate:
High:
Complexity Joint encoding between
Decoding at each relay.
source and relays.

CSI Required Yes No

Latency nt+m ntm


(Coding block length)

nt, nr : the number of antennas at S and D


m: the number of relay stations

2007 Robert W. Heath Jr. 15


2007 Huawei May 2007

Key Results on DSTC (2/3)

Diversity and Multiplexing Tradeoff for Repetition and DF-STC

m: the number of DF relays

Rnorm: spectral efficiency

(Source: [Laneman et al., IT03] )

When R0, DF-DSTC and repetition strategies provide the full diversity order m
DF-DSTC strategy provides much more spectral efficiency than repetition-strategy,
especially as m becomes large

2007 Robert W. Heath Jr. 16


2007 Huawei May 2007

Key Results on DSTC (3/3)

(Rnorm) Diversity and Multiplexing Tradeoff for Opportunistic AF

1 ideal

DF-STC

Opportunistic AF m: the number of DF relays

Rnorm: spectral efficiency


Repetition

Non-cooperation
(Source: [Bletsas, Ph.D 05])
Rnorm
1/m 1/2 1

The diversity-multiplexing of opportunistic amplify-and-forward is exactly


same as that of decode-and-forward space-time code (DF-DSTC)

2007 Robert W. Heath Jr. 17


2007 Huawei May 2007

Collaborative Spatial Multiplexing


MS/RS

collaboration?
BS
MS/RS
Spatial Multiplexing Decoder
Collaborative SM Encoder

For uplink, adopted in IEEE802.16e (Sec. 8.4.5.4.1.9)


Requires minimal coordination between users
High capacity solution,not useful for diversity/coverage
Primarily for mobile RS but can be used with fixed RS

2007 Robert W. Heath Jr. 18


2007 Huawei May 2007

Case Study of Collaborative SM (1/3)


L hops

x1 x1,x2

R R R
x1,x2
x1 orthogonal
transmission
S H1 HL-2 D
concurrent
x2 transmission
x1,x2
R R R
x2 x1,x2 (Source: [Wittneben et al., Assilomar 05])

Multi-hop with amplify-and-forwad (collaborative SM)


- D should know all channel state information about H1 through HL-1
- Each relay transmit via concurrent channel except for the last hop

2007 Robert W. Heath Jr. 19


2007 Huawei May 2007

Case Study of Collaborative SM (2/3)

L hops

x1 x1

R R R
x1
x1
orthogonal
S D transmission

x2
x1
R R R
x2 x2 (Source: [Wittneben et al., Assilomar 05])

Multi-hop with decode-and-forward (Non-collaborative SM)


- Each relay decodes the received signal
- Transmit all signal via orthogonal channel (TDMA, FDMA, etc.)

2007 Robert W. Heath Jr. 20


2007 Huawei May 2007

Case Study of Collaborative SM (2/3)

collaborative MIMO

normalization

non-collaborative MIMO

Normalization is required because DF exploits orthogonal


transmission
Collaborative MIMO outperforms non-collaborative MIMO

2007 Robert W. Heath Jr. 21


2007 Huawei May 2007

MIMO Relay

RS

BS MS

BS, MS, and possibly RS have MIMO capability


Uses direct and relay paths
Can use beamforming or DPC message splitting
Requires coordination and CSI exchange between BS, RS, MS
Capacity region is still unknown.

2007 Robert W. Heath Jr. 22


2007 Huawei May 2007

Theoretic Results on MIMO Relay (1/5)


Cover and El Gamal (IT 79)
- Rate bounds for general full-duplex relay channel
Wang, Zhang and Host-Madsen (IT 05)
- Rate bounds for Gaussian MIMO relay channel
- Non-cooperative lower bound for fixed channels

Relay H3
H1

Sender Receiver
H2

2007 Robert W. Heath Jr. 23


2007 Huawei May 2007

Theoretic Results on MIMO Relay (2/5)


X1 is split into U and V
Relay treats V as interference

Relay H3
H1
X1 = X1(U,V) X2 = X2(U)

X1 = X1(U,V)
Sender Receiver
H2

2007 Robert W. Heath Jr. 24


2007 Huawei May 2007

Theoretic Results on MIMO Relay (3/5)


Superposition coding
- X1 is linear combination of independent U and V
- Rx decodes U and V while relay just decodes U

Relay
X1 = U + V X2 = X2(U)

X1 = U + V
Sender Receiver

2007 Robert W. Heath Jr. 25


2007 Huawei May 2007

Theoretic Results on MIMO Relay (4/5)


Dirty-paper coding
- Tx chooses U based on known interference V

(Source: [Peel et al., Sig. Proc. Mag. 03] )

2007 Robert W. Heath Jr. 26


2007 Huawei May 2007

Theoretic Results on MIMO Relay (5/5)


Message splitting outperforms Wang et al. lower
bound
- Wang et al. lower bound ignores relay path or direct path

(Source: [Lo, Vishwanath and Heath, Globecom 05])

2007 Robert W. Heath Jr. 27


2007 Huawei May 2007

Multiuser MIMO Relay*

MS
BS RS

Use MIMO broadcast from RS to MS to support multiple users


Use MIMO between BS & RS to achieve higher capacity
Provides higher sum capacity and better coverage
Use DPC type encoding at BS for best performance
Requires coordination and CSI exchange between BS, RS, MS

* Proposed by Samsung & UT (Source: [Tang, Chae and Heath, appear in SP06] )

2007 Robert W. Heath Jr. 28


2007 Huawei May 2007

Concept of Virtual Antenna Arrays

Generic VAA scheme


Virtual Antenna Array (VAA) Cell

Co-located users form VAA


Users in VAA collaborate to forward data
Several STC-based approaches proposed
Coordination protocols, synchronization overhead

2007 Robert W. Heath Jr. 29


2007 Huawei May 2007

Two-way Relaying (Bi-Directional) Communication


p1: Phase I, p2: Phase II, p3: Phase III, p4: Phase IV

(p2) x3=x1 + x2
(p4) x4=x2 (p3) x3=x1
R R
(p1) x1 (p2) x2 (p1) x1 (p1) x2
S D S D

(a) Conventional Relay Channel (Routing) (b) Two-way relaying (Network coding)

Different strategies at relay


- Decode and forward (digital network coding)
- Amplify and forward (analog network coding)
- Compress and forward

2007 Robert W. Heath Jr. 30


2007 Huawei May 2007

Two-Way Relaying (Contd)


Advantages
- Overcomes half-duplex restriction

Capacity is known for decode and forward when S and


D encode independently [Boche et al., ICASSP 07]
[Tarokh et al., IT submitted]
- min(CMAC(phase 1), CBC(phase2)) with half-duplex relay
MAC: multiple access channel
BC: broadcast channel

2007 Robert W. Heath Jr. 31


Relay Solutions of WSIL

- [S1] Multi-user MIMO with a fixed relay


- [S2] Relay Selection
- [S3] Capacity of two-way relaying with large number of relays
2007 Huawei May 2007

[S1] MIMO Solutions for Fixed Relay


Poor coverage for some users in the downlink
- Users are located at the cell edges or under shadowing
- Opportunistic scheduling may not solve this problem

Using a fixed relay BS


NK!" - Design option in 802.16j
- Practical considerations
Low complexity relays [Pabst et al04]
MS
Easy to configure, compatible to
a downlink system without relays RS

Proposed design option


- Use a MIMO relay with linear processing at the relay

(Source: [Tang, Chae and Heath, appear in SP06] )

2007 Robert W. Heath Jr. 33


2007 Huawei May 2007

Proposed Multiuser MIMO Relay


MS
BS RS

Use MIMO broadcast to support multiple users


- Leverage results from [CaireShamai03][Vishwanath et al03]
[YuCioffi02][ViswanathTse03]

Use MIMO between BS & RS to achieve higher capacity


- Leverages MIMO capacity results [Foschini][Telatar]

Application: coverage enhancement, multiple voice users

2007 Robert W. Heath Jr. 34


2007 Huawei May 2007

[S1] MIMO Relay Broadcast

Base Station (BS) Relay Station (RS)

MS 1

Encode F W
MIMO Link
MS 2
DL signalling/
Optimization/ RS scheduling
scheduling
MS Nu
Feedback MIMO BC

Use linear processing at the MIMO relay to serve multiple users


Propose a relaying structure with Tomlinson-Harashima
precoding

2007 Robert W. Heath Jr. 35


2007 Huawei May 2007

Proposed System w/ Linear Processing 2/2


Assumptions
- Neglect direct link between BS & MS
- Two-hops (MIMO link + MIMO broadcast), AWGN
- Each MS has one receive antenna
- Only linear processing allowed at the RS
- Channel state globally known (neglect feedback, errors)

Signal model for the ith user

Power constraints

2007 Robert W. Heath Jr. 36


2007 Huawei May 2007

Sum Rate Lower Bounds 1/3


Derive bound w/ DPC at BS + certain structure on W
SVD relay design
- Define the SVD and QR decompositions

- Define the transmit precoder and relay precoder

- Where and K are diagonal power allocation matrices with


diagonals p and k

2007 Robert W. Heath Jr. 37


2007 Huawei May 2007

Sum Rate Lower Bounds 2/3


Write effective SNR for ith user of permutation

Sketch of SVD relay sum rate bound derivation


- Write sum rate expression based on effective SNR
- Lower bound sum rate
- Write constraints in terms of vecs p and k
- Maximize subject to power constraints
- Theorem: Resulting optimization is a geometric program
(follows from posynomial cost and inequality constraints)
Can solve using convex techniques via transformation of variables

Must optimize over all possible user permutations

2007 Robert W. Heath Jr. 38


2007 Huawei May 2007

Sum Rate Lower Bounds 3/3


Equal power allocation lower bound
- BS selects first T Mu = min (Mr, Mb, Nu) users
- Equal power allocation for T users at BS for vector p
- Leads to a lower bound on achievable rate

Greedy reduced complexity search algorithm


- Use lower bound to derive user selection (permutation) alg
- Based on results from [TuBlum03] [DirmicSidiropoulos05]

After user selection, perform waterfilling at relay for


better performance

2007 Robert W. Heath Jr. 39


2007 Huawei May 2007

Sum Rate Upper Bound


A sum rate upper bound for the proposed system
- Capacity C1(H1) of the MIMO link found via the usual waterfilling
- Sum capacity C2(H2) of the second link found via your favorite MIMO BC
channel solution like duality

Achievable rate of decode & forward w/ optimal time sharing

2007 Robert W. Heath Jr. 40


2007 Huawei May 2007

Comment on Practical Details 1/2


Channel info. of BS->RS Channel info. of RS-> MS

Data/Control Link 1 Data/Control Link 2

BS RS MS
Feedback link 2 Feedback link 1

Case 1: BS performs optimization/scheduling


- Feedback of the channel info. of RS MS from MS to BS via
RS and the channel info. of BS RS from RS to BS
- Send optimum W to RS
Case 2: RS performs optimization/scheduling
- Feedback of the channel info. of RS MS from MS to RS
- Send optimum F from RS to BS
Reciprocity / uplink channel sounding can simplify
2007 Robert W. Heath Jr. 41
NK!"
2007 Huawei May 2007

[S1] A Precoding Strategy


Propose to adaptively select the number of streams and QAM modulation
NK!" - Select a group of users and determine transmission strategies with SER targets
- Use the Tomlinson-Harashima precoding (THP) at the base station
- Select a subset of users based on their channels (a user selection algorithm)
- Determine the QAM order based on the channel gain of each stream

User selection
Mod Derived a SNR lower bound
AMC F
B-I Power allocation

Optimization
Decide QAM
Base Station

Channel Information of Users


Any zero rate?
No
Yes Terminate
Drop the stream

2007 Robert W. Heath Jr. 42


NK!"
2007 Huawei May 2007

[S1] Bounds of Sum Rates

This MIMO relay channel resembles a broadcast channel


NK!"

Sum rate lower bounds


- Derive bounds with dirty paper coding W
and structure the relay processing matrix

! k1 "
W = U ## k2 $V
$
#% k3 $&

SVD relay

A sum rate upper bound


- Half of the capacity of the
first hop MIMO channel

2007 Robert W. Heath Jr. 43


2007 Huawei May 2007

[S1] System Comparison


RS Operation Items Proposed System Concatenated System

Matrix Multiplication Yes Yes

Matrix Inversion No Yes


(MMSE / ZF)
Error decoding No Yes
(e.g.,Turbo decoding) (high complexity)

Multiuser BC encoding No Yes


(e.g.,Dirty paper coding) (BS takes this (high complexity)
functionality)
Scheduling Maybe Yes
(user selection and adaptive (May use additional (MAC function)
modulation) scheduling at the BS) (high complexity)

2007 Robert W. Heath Jr. 44


2007 Huawei May 2007

[S1] Simulation Results

2007 Robert W. Heath Jr. 45


2007 Huawei May 2007

Average Sum Rate (bps/Hz) [S1] Benefits of Relays

! =1 , dist(BS-MS):dist(RS-MS)=5:5
! = 1.5 , dist(BS-MS):dist(RS-MS)=6:4
! = 2.3 , dist(BS-MS):dist(RS-MS)=7:3

} Proposed SVD relay approaches *

Sum capacity without Relay Station


Distance between BS and MS

Cell coverage and sum rate enhancement


- Cell radius gain = 300m @ target sum rate = 4 bps/Hz
- Sum rate gain = 2bps/Hz @ cell radius = 1km

2007 Robert W. Heath Jr. 46


2007 Huawei May 2007

[S2] Relay Selection (1/2)


Hybrid-ARQ with relays yields higher throughput
Which relays forward parity or resend packet?

Strong Strong
link Relay link

ARQ
Source Error Sink
Weak link
Strong Strong
link Relay link

(Source: [Lo, Heath, Vishwanath, VTC 07])

2007 Robert W. Heath Jr. 47


2007 Huawei May 2007

[S2] Relay Selection (2/2)


De-centralized relay selection strategy
Hybrid-ARQ supported via parity bit forwarding

Transmit
Source Tx
Listen Listen Chosen Listen
Framing Rate-R1 Relay ID

Hello
Messages

Chosen Relay
Tx Rate-R2
Relay Listen Listen Listen
Framing Other Relays
Listen

Contention
Minislots
Sink Tx
Framing Listen ACK or Listen Listen Listen
NACK

2007 Robert W. Heath Jr. 48


2007 Huawei May 2007

[S2] Simulation Results (1/2)


Opportunistic selection outperforms direct transmission that
ignores relays
Opportunistic selection also competitive with GPS-based strategy

2007 Robert W. Heath Jr. 49


2007 Huawei May 2007

[S2] Simulation Results (2/2)


Use extra bit of channel feedback to refine proposed selection
strategy
Limited feedback closes gap with best gain approach that
always picks relay with best gain to rx

2007 Robert W. Heath Jr. 50


2007 Huawei May 2007

[S3] Capacity Scaling (Two-way Relaying)

Phase I Phase II

1 1
R R
x1 x2 x1 + x2 x1 + x2
2 2
S R D S R D
x1 x2 x1 + x 2 x1 + x2

x1 x2 x1 + x 2 x1 + x2
m m
R R

(Source: [Vaze and Heath, ISIT 07])

2007 Robert W. Heath Jr. 51


2007 Huawei May 2007

[S3] Capacity scaling continued


With Large number of relays, the sum capacity scales as ntlog(m)
(where m are number of relays and nt are the number of antennas
at S and D).
In the limit as number of relays go to infinity, for MIMO Gaussian
channels, capacity is given by (Vaze and Heath)

R1
(nt log(m), nt log(m))
nt log(m)

nt log(m) R2

where R1 and R2 are rates from S to D and D to S respectively.

2007 Robert W. Heath Jr. 52


2007 Huawei May 2007

[S3] Two-way relaying advantage


Improvement in capacity by a factor of 2 as number of relays grow
large

2007 Robert W. Heath Jr. 53


2007 Huawei May 2007

Summary and Future Work


MIMO multiuser relay w/ linear amplify & forward
- Lower bounds on sum rate
- Algorithms for finding transmit and relay precoders
- Algorithm for user selection
- Upper bound based on decode and forward

Future work
- Capacity bounds including the BS to MS connection
- Limited feedback, quantized precoding, limited feedforward

2007 Robert W. Heath Jr. 54


2007 Huawei May 2007

Summary
Cooperation is the next phase of cellular development

Cooperation entails new levels of coordination


UT is developing promising cooperative technologies
- Linear MU-MIMO achieving near sum capacity (IEEE802.16m)
- Adaptive MIMO Handover (IEEE802.16j/m)
- Enhanced Amp and Forward Relays (IEEE802.16j).
High Performance and Low Cost

2007 Robert W. Heath Jr. 55

Potrebbero piacerti anche