David Tse Wireless Foundations Dept of EECS UC Berkeley CWIT May 20 2011 TexPoint fonts used in EMF Read the TexPoint manual before you delete this box A A A A A A A A ID: 344786
Download Presentation The PPT/PDF document "Feedback for Interference Mitigation" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Feedback for Interference MitigationDavid TseWireless Foundations Dept. of EECS U.C. BerkeleyCWIT May 20, 2011
TexPoint fonts used in EMF.
Read the TexPoint manual before you delete this box.:
A
A
A
A
A
A
A
A
A
ASlide2
2 Drive for Wireless Spectral Efficiency Lots of PHY layer advances in the past 1.5 decade.Focus on point-to-point and
single-cell performance.Recent research has shifted to interference.Slide3
The Interference BarrierInterference gets worse as:density of wireless nodes increases.wireless architectures become more heterogeneous and decentralized
Heterogeneous networks
Macro-cell
pico-cell
Tx1
Rx1
Peer-to-peer networks
Tx2
Rx2
Tx3
Rx3
3Slide4
What I’ll Talk AboutDiscuss several interference mitigation approaches.Focus on a common key enabler: feedback.Slide5
Interference Mitigation ApproachesMultiuser MIMONetwork MIMOInterference alignment5Slide6
Downlink Multiuser MIMO mitigates intra-cell interference
6
Can transmit K symbols/s/Hz
using K transmit antennas.
(degrees of freedom = K)Slide7
Inter-Cell Interference Mitigation Multiuser MIMO can also mitigate
inter-cell interference
7Slide8
Network MIMOmitigates inter-cell interference 8
Central
UnitSlide9
Inter-cell Interference Mitigation With multiuser MIMO, each out-of-cell user to be nulled out costs one dimension. Network MIMO removes that cost at the expense of additional infrastructure. Can one reduce that cost without infrastructure cooperation? Slide10
Example: Single Tx Antenna at BS10Cannot beamform with one antenna.D.o.f. per cell = ½ sym/s/Hz (frequency sharing)Turns out that with more users per cell and frequency diversity, one can do better.
X
XSlide11
Downlink Interference Alignment
Interference alignment between
out-of-cell
and
intra-cell
interference
11
Fix 2 dim. reference
plane in 3-dim signal space,
independent
of channel gains
2
1
1
1
2
2
transmission over
3 sub-carriers
(
Suh
et al 10)Slide12
Key Enabler For All Techniques Timely channel knowledge at the transmitters: Multiuser MIMONetwork MIMO
Interference alignment.
12
Central
Unit
2
1
1
1
2
2Slide13
Channel State FeedbackSuch CSI is typically obtained via feedback.Feedback overhead:f = 2 GHz, v = 30m/s, delay spread 5 m s coherence time ¼ fv/c = 5mscoherence bandwidth ¼ 200 kHz coherence block ¼ 1000 sym
Feedback delayseveral millisecondscan be of order of channel coherence time
Delay
is a critical issue in hi-mobility.
Coherence
blockSlide14
Conventional ApproachTransmitter predicts current channel state based on fed back information.Predicted channel state is used in place of true state in interference mitigation schemes.Slide15
Prediction is very difficult, especially if it’s about the future. Niels Bohrs (or was it Yogi Berra?)15Slide16
Completely Outdated FeedbackWe ask: What if current channel is completely independent from the fed back information?Conventional wisdom: Feedback is totally useless.Is conventional wisdom correct?
16Slide17
Downlink MIMO Perfect channel knowledge:
2 symbols per time slot
A
B
17Slide18
Downlink MIMO Perfect channel knowledge: 2 symbols/s/Hz No channel knowledge:
1
symbol/s/Hz
18Slide19
Downlink MIMO Outdated channel knowledge?4/3 symbol/s/Hz
reconstruct
19Slide20
K by K Downlink MIMO
symbols/s/Hz
K antennas
K
receivers
(Maddah-Ali & T. 2010)
Perfect CSI: K symbols/s/Hz
No feedback:
1 symbol/s/Hz
Outdated
feedback: Slide21
3 x 3 Case
Phase I
Transmit:
Symbols for
individual
users
Symbols wanted by
2
users
Phase II
Phase III
Symbols wanted by
3
usersSlide22
Details on Phase II
Phase III:
Transmit any
two
linear combinations of
the threeSlide23
Information Theoretic OptimalityTheorem: The d.o.f. of the K by K MIMO broadcast channel with i.i.d. Rayleigh fading under feedback is:Outer bound:Physically-degraded BC:Feedback does not help (El Gamal 78)Slide24
Extension to Interference Channels
reconstruct
24
2-user
MIMO IC with arbitrary # of antennas:
(
Vaze
and Varanasi 11
)
3-user IC (
Maleki
et al 10)Slide25
Feedback in Information Theory Shannon 1956:
Feedback does not increase capacity of
memoryless
channels.
X
memoryless
channel
encoder
decoder
memoryless
channel
encoder
delay
decoderSlide26
Feedback in PracticeHybrid ARQ/ rateless codes:Tries to achieve the capacity without knowing h.Prediction Tries to exploit memory in channel process {h[m]}.
encoder
decoder
encoder
delay
decoder
encoder
decoder
encoder
delay
decoder
unknown hSlide27
Feedback in Memoryless Networksencoder
dec 1
dec
2
memoryless
broadcast
channel
delay
delay
Dueck
provided an example of a BC where feedback helps
but the noises at the two receivers are dependent.
Our result:
feedback increases not only capacity but
d.o.f
.
even when noises at receivers are
independent
.
What is this
new
role of feedback? Slide28
Side Information
reconstruct
28
side info.
side info
.Slide29
Role of FeedbackOld:Predict channel.New:Learn about side information at receivers.29Learning about the past is often easier than predicting the future.Slide30
ConclusionsInterference is a central barrier to scalability of wireless systems.Multiple approaches to mitigate interference are emerging.New use of feedback is a key enabler for these approaches.30Slide31
Reference M. Maddah-Ali and D. Tse, “Completely Stale Transmitter Channel State Information is Still Very Useful”, Allerton Conference, 2010. http://arxiv.org/abs/1010.149931