The Practical Challenges of Interference Alignment

The Practical Challenges of Interference Alignment

Daniel Tai

10/28/2013

Daniel Tai The Practical Challenges of Interference Alignment 10/28/2013 1 / 18

Reference

1 O El Ayach et al., “The Practical Challenges of Interference Alignment”, IEEE WirelessCommunications, Issue 1, Vol. 20, 2012


Intoduction to Interference Alignment

Outline

1 Intoduction to Interference Alignment

2 Challenges of IA

3 Solutions

4 Some Numerical Results



Interference Alignment

The idea of interference alignment (IA) is to use channel’s multiplexing gain (degree offreedom) to cancel out interference at different users.

Specifically, IA allow users to cooperatively design precoder/decoder to project (“align”)all interferences onto the same space different than signal space.



IA Formulation

To be even more specific,

Consider a system with K users with Si datastreams to transmit to each.

Transmitting symbol si ∈ CSi×1 is encoded using precoder F

Hi ,j denotes channel matrix from the system intended for user j to user i

Received signal at user i :

yi = Hi ,iFisi +∑l 6=i

Hi ,lFlsl + ni

where ni is noise

Hi ,j and Fi ’s dimensions and structures depend on the type of systems (for example Hi ,j

is diagonal for TDD/FDD system)

The goal of IA is to design Fi1 such receivers are able to cancel out all interferences. (For

example, if linear decoder Wi is used, W?i Hi ,lFl = 0∀l)

1(my interp.) decoders might as well, but this paper focuses on precoder designDaniel Tai The Practical Challenges of Interference Alignment 10/28/2013 5 / 18

Challenges of IA

Outline


2 Challenges of IA

3 Solutions



Challenges of IA

Challenges of IA

Dimensionality and ScatteringThe # of dim. need for IA grows exponentially with # of users.Relatively milder for MIMO systems which we can add antennas(This paper mainly focuses on MIMO IA)

SNRIA performs well at high-SNRIn moderate-SNR, IA may not reach theoretical channel capacity. (discussed later)

CSI Estimation and FeedbackCSI need causes overhead/quality trade-off and other performance degradations intransmission.

SynchronizationIA technique needs tight synchronization (no CFO, CTO) between cooperating nodes.Insufficient synchronization causes additional noise.Currently GPS solution is used.

Network OrganizationNodes needs to share system parameters, CSI... etc.


Solutions

Outline


2 Challenges of IA

3 Solutions



Solutions

Computing IA Solutions I

Iterative algorithms are generally used to compute IA

Earliest method 2:

At each iteration, leakage is minimized.Ideally, W?

i Hi,lFl = 0 at convergence point.Good performance in high-SNR.Problem: Oblivious to desired signal power, so it performs far from optimal in low-SNR orbad channels.

Improvements:2 Changing the goal to max. per-stream SINR. (Performs also well in low-SNR)Maximizing sum rate3: The paper gives the equivalence between max sum rate and min sumMSE. Hence MMSE solution is developed.

Other limitations

Allows users to cooperate generate lots of uncoordinated (colored) noise:Algorithm enhanced in 4


Solutions

Computing IA Solutions II

CSI sharing might causes large overhead:5 uses game theory to replace matrix feedback by scalar feedback.

2K. Gomadam, V. Cadambe, and S. Jafar, “A Distributed Numerical Approach to Interference Alignmentand Applications to Wireless Interference Networks,” IEEE Trans. Info. Theory, vol. 57, no. 6, June 2011, pp.3309-22.

3Q. Shi et al., “An Iteratively Weighted MMSE Approach to Distributed Sum Utility Maximization for aMIMO Interfering Broadcast Channel,” IEEE Trans. Signal Proc., vol. 59, no. 9, Sept. 2011, pp. 433140.

4S. W. Peters and R. W. Heath, Jr., “Cooperative Algorithms for MIMO Interference Channels,” IEEETrans. Vehic. Tech., vol. 60, no. 1, Jan. 2011, pp. 206218.

5C. Shi et al., “Local Interference Pricing for Distributed Beamforming in MIMO Networks,” Proc. IEEEMILCOM, Oct. 2009.


Solutions

Obtaining CSI I

2 Mainstream methods: Channel reciprocity and feedback


Solutions

Obtaining CSI II

Problems of reciprocity method

Iterative causes overhead (uses lots of timeslots)

Reciprocity might not stand.(ex. uncoordinated interference observation at rx and tx arenot reciprocal)

Does not work for FDD systems

Problems of feedback method

Quality(accuracy)/overhead tradeoff

Limited feedback is low-overhead, but efficient quantization such as Grassmanniancodebooks grows exponentially with accuracy requirement in high SNR ⇒ hard togenerate and encode.

Limited feedback cannot be applied to systems without structured CSI.

Analog feedback: prone to thermal noise (bad for uplink)


Solutions

Obtaining CSI III

Other problems

CSI data grows with network size: overhead may cancel possible gain

BS backhaul link requirement: high capacity and low latency


Solutions

IA in large scale networks

6 Shows that in partially connected networks, a finite number of antennas can be used toIA in infinitely large network.

Issues: Realistic? How to set the threshold of interference?7 has a more complicated model: a finite channel coherence time and different path lossto each link. CSI acquisition is also considered.⇒ TDMA performs better for fast fading channels. The paper also providing partitioningalgorithm to make IA/TDMA hybrid systems.

6M. Guillaud and D. Gesbert, “Interference Alignment in the Partially Connected K-User MIMO InterferenceChannel,” Proc. Euro. Signal Proc. Conf., Barcelona, Spain, Sept. 2011, pp. 15.

7S. W. Peters and R. W. Heath, Jr., “User Partitioning for Less Overhead in MIMO Interference Channels,”IEEE Trans. Wireless Commun., vol. 11, no. 2, 2012, pp. 592603.


Some Numerical Results

Outline


2 Challenges of IA

3 Solutions




Effects of Limited Scattering



Effects of CSI Mismatch



Effects of User Grouping


Date post:	06-May-2015
Category:	Technology
Upload:	daniel-tai
View:	377 times
Download:	0 times

The Practical Challenges of Interference Alignment

Technology