Roma, February 3, 2006WOMEN Project - Kickoff meeting1 PRIN 2005 WOMEN Project– Kickoff meeting Research Unit Università of Napoli Federico II Activities

Roma, February 3, 2006 WOMEN Project - Kickoff meeting 1

PRIN 2005 WOMEN Project– Kickoff meeting

Research Unit Università of Napoli Federico II

Activities to be carried out

within the first semester

Giacinto Gelli

DIET, Università of Napoli Federico II

[email protected]


Research Unit in brief

• Manpower: 6 researchers + 1 postdoc + 1 PhD student

• Leader of WP2 “Multi-antenna transceivers for mesh

networks” encompassing the following tasks:– T2.1 “Multi-carrier space-time modulation and multi-antenna coding

techniques for broadband fading channels” (Roma, Napoli);

– T2.2 “Efficient receiver multi-antenna architectures” (Napoli);

– T2.3 “Smart antenna techniques for alien-interference mitigation”

(Napoli);

– T2.4 “Adaptive beamforming” (Roma).

• Other research activities:– T3.6 “Mesh connectivity layer” within WP3 coordinated by Firenze.


Transceiver design for MIMO channels

• Task T2.1

• Deals with multiple input/multiple output (MIMO) communication systems.

• MIMO topics:– multiple antenna systems;

– multiuser detection;

– multicarrier systems (es. OFDM, MIMO-OFDM);

– diversity techniques.


Goals of MIMO research

To design communication links that offer:

High data rate

Maximize the channel capacity

Quality of service

Minimize the error rate

Low-cost implementation

Trade-off between performances and computational complexity


MIMO channel model

• Assumptions for the lowpass representation:– channels modeled as finite impulse response (FIR) linear time-

invariant (LTI) systems;

– additive noise.

No

ky

1kn

1kx

2kx

iNkx

1ky

2ky

1,1kh

1, oNkh

1,2kh 2

kn

oNkn

,i oN Nkh

2,2kh

,1iNkh

0k m k m k

m

y H x n

1,1,1 1,2

2,1 2,2 2,1

,1 ,2 ,

i

o o o i

Nk k k

k k kk

N N N Nk k k

h h h

h h h

h h h

H

inputs

iNoutputs

oN


MIMO equalization (1/2)

2y

3y

1x

2x

3x

1y

1n

2n

3n

Receiver?

Research topics (under study):

• MMSE equalization techniques exploiting the statistical redundancy due to non-circularity of the channel input.


MIMO equalization (2/2)

2y

3y

1x

2x

3x

1y

1n

2n

3n

Receiver?

Proposed structures:

• Widely linear (WL) FIR receivers based on the decision feedback strategy.

MIMO channel

WL feed-forwardfilter Decision device

WL feed-backfilter

x x̂


MIMO transceivers

2y

3y

1x

2x

3x

1y

1n

2n

3n

Transmitter (TX)

?

Receiver(RX)

?

Channel State Information (CSI)

Furher research topics:

• Joint design of the TX and RX with transmit power constraint:– WL-FIR TX/RX;

– optimization critera: MMSE, mutual information, QoS.


Blind adaptive channel shortening (1/3)

• Task T2.2

• In multicarrier systems, channel frequency selectivity can be compensated for by inserting a cyclic prefix (CP) longer than the channel impulse response (CIR).

• Highly time-dispersive channels -> long CP -> significant reduction of channel throughput.

• Goal: Minimize throughput reduction by means of a time-domain equalizer (TEQ) at the receiver front-end:

– the TEQ shortens the channel so that the combined channel-equalizer

impulse response is shorter than the CP length.



• Traditional channel shortening (CS) techniques need channel knowledge or training sequences. – Drawback: the use of training sequences reduces channel

throughput !

• Blind CS approaches are able to shorten the CIR, without requiring training sequences.

• A blind CS algorithm must exhibit three desirable features:

– suitable for a large class of CIR;

– manageable complexity (adaptive implementation);

– fast and global convergence.



• Existing blind CS techniques rely on:

– CP redundancy: • low complexity, global convergence; • a large amount of data is required to converge.

– Auto-correlation minimization:• fast convergence, tracking capabilities; • high complexity, global convergence is not ensured;

– Oversampling of the received signal:• fast convergence, high performance;• high complexity, batch processing (non adaptive); • restrictive assumptions on the CIR to be shortened.


Proposed approach (1/2)

• The proposed approach (under investigation) relies on oversampling the received signal.

• Thanks to the time redundancy induced by oversampling, the channel convolution matrix exhibits quite a rich structure, since each column can be linearly parameterized as

where is a known matrix, whereas collects the unknown channel parameters.

• The number of columns of depends on the length of the CIR to be shortened.

H

n n nh Q c

nQ nc

H


Proposed approach (2/2)

• The combined channel-equalizer response is

where collects the TEQ parameters. Shortening the CIR amounts to force to zero some entries of .

• Vector is chosen so as to minimize the mean-output-energy (MOE) at the TEQ output, with blind constraints preserving only a small number of entries of (smaller than the CP length).

• Blind constraints are imposed by resorting to the aforementioned parameterization of .

• Expected features of the MOE-based approach:– easy adaptive implementation;– fast convergence;– mild conditions on the CIR to be shortened.

Hf H gg

f

g

f

H


Equalization, channel identification, and NBI suppression

• Task T2.3

• Space-time block coding (STBC) exploits both temporal and spatial diversity, enabling a significant increase in transmission rate.

• To decode STBC, channel state information (CSI) must be acquired at the RX by training or blind methods. The amount of training data increases with the number of TX and RX antennas.

• To avoid a throughput decrease, training approaches can be integrated with blind ones (semi-blind approach), shortening thus the training period.

• Equalization and interference suppression in multiantenna systems is also more challenging than in single-antenna systems.


Alamouti’s STBC (1/3)

• Due to size and power limitations, mobile units usually cannot employ more than two antennas.

• Alamouti’s STBC (AL-STBC) is a popular and practical technique employing two transmit antenna and one receive antenna.

• In AL-STBC, two consecutive symbol blocks and are subject to space-time encoding:

*

*

(2 ) - (2 1)(2 1) (2 )

n nn n

s ss s

( )ns (2 1)ns

space

time



• AL-STBC originally proposed for flat-fading channels:– maximum-likelihood (ML) decoding can be performed by using

linear processing and multiantenna diversity of order two can be achieved.

• AL-STBC can be generalized to frequency-selective channels:– ML decoding is computationally heavy: simple linear ML decoding

is not directly applicable.

• To maintain decoding simplicity and advantages of AL-STBC for flat-fading channels, suboptimal decoding approaches must be pursued over frequency-selective channels.



• AL-STBC can be regarded as a widely-linear (WL) precoding, which generates an improper (non circular) transmitted signal.

• When the transmitted signal is improper, it is well-known that WL processing is beneficial.

• Existing suboptimal decoding approaches for AL-STBC multicarrier systems rely on linear processing and thus do not fully exploit the improper nature of the transmitted signal.


Proposed approach

• WL processing of the received data:– the received signal and its complex conjugate are jointly elaborated;

– the dimensionality of the observation space is doubled -> additional degrees of freedom for RX synthesis.

• Research topics:– Synthesis of WL generalized zero-forcing (ZF) equalization

structures with NBI suppression capabilities.

– Design of improved (semi)-blind methods for acquiring CSI.


Mesh connectivity layer (1/2)

• Task T3.6

• Goals: To design a functionality for network topology monitoring and to identify its possible implementation in order to:

– provide to any active node the topology knowledge, in terms of

node position, state and connectivity;

– support routing and network management.


Mesh connectivity layer (2/2)

• The mesh connectivity layer is composed of:

– A cooperative and distributed mechanism for connectivity management in the backbone layer.

+– A cooperative and distributed mechanism for localization and

mobility management in the ad hoc layer.

• In both domains the hierarchic organization of the network has to be exploited.


Ad hoc domains

• The main problem consists of localization and mobility management.

• A procedure for network clustering (each cluster controlled by a single Wireless Router - cluster head) is necessary.

• A procedure for monitoring, acquisition and distribution of localization information inside the cluster and toward the cluster head (WR) is necessary.


Localization management

• The solution is based on a virtual backbone to support

localization-information distribution inside the network and

toward the cluster head.

• We look for a distributed algorithm for virtual backbone

building and dynamic updating.


Virtual backbone

• The virtual backbone identifies a subset of nodes as

Location Servers (LS).

• The virtual backbone must:– minimize the number of LS;

– control the overhead level;

– exploit the intrinsic hierarchic organization of the network.


Application

• The mesh connectivity layer provides a mechanism for

network topology control via cross-layer approach.

• It can be exploited to support routing, but also for a new

MAC protocol for wireless mesh networks.


MAC research activity

• Solution: to use the already existent 802.11 MAC protocol and to design an LLC level able to manage temporal and frequential multiplexing and to control the topology in order to adaptively assign the transmitting resources, both reducing the collisions and increasing the throughput.

• The mesh connectivity layer provides updated information about the network topology and state, and allows one to design a highly adaptive MAC protocol.

Documents

Roma, February 3, 2006WOMEN Project - Kickoff meeting1 PRIN 2005 WOMEN Project– Kickoff meeting Research Unit Università of Napoli Federico II Activities