PHY-MAC Dialogue with Multi-Packet Reception

PHY-MAC Dialogue PHY-MAC Dialogue with Multi-Packet Receptionwith Multi-Packet Reception

Workshop on Broadband Wireless Ad-Hoc Networks and Services

12th-13th September 2002ETSI, Sophia Antipolis, France

Marc Realp-CTTC/Ana I. Pérez-Neira-UPCwww.cttc.es www-tsc.upc.es

IST-2001-38835 ANWIRETIC2002-04594-C02-02 GIRAFA

ContentsContents

Motivation Cross-Layer Design

MPR matrix. PHY-MAC dialogue. Parameters Exchange.

PHY level Matched Filter. Activity User Detection.

MAC level Dynamic Queue Protocol-DQP. Modified Dynamic Queue Protocol-

MDQP. Simulations Conclusions & Further Work

MotivationMotivation

In wireless systems a common channel is shared by many users.

Traditionally, information is lost when a collision occurs, i.e., when two or more packets are sent trough the channel.

Diversity at physical level allows more than one packet to be transmitted simultaneously.

Conventional MAC algorithms do not consider Multi-Packet Reception (MPR) capability.

CROSS-LAYER DESIGNMAC fully exploits PHY reception capabilities.

Motivation Cross-Layer

Design PHY level MAC level Simulations Conclusions &

Further Work

MPR matrixMPR matrix

The probability of a packet to be correctly received is:

Hence,

,

1 (num. of users) ; 0

[ packets are correctly received | packets are transmitted]

( , , ( ))n k

n M k n

C P k n

B k n Ps n

0

Pe(n) Bit error probability in the presence of 1 interferers

e=number of correctable errors in a packet

pl=packet length

( ) ( , , ( ))e

t

n

Ps n B t pl Pe n



Further Work

MPR matrixMPR matrix

The MPR matrix is defined as:

Expected number of correctly received packets when n packets have been transmitted:

1,0 1,1

2,0 2,1 2,2

,0 ,

0 0

0

0

M M M

C C

C C C

C

C C

,1

n

n n kk

C kC



Further Work



Further Work

PHY-MAC DialoguePHY-MAC Dialogue

Cross-Layer design reduces PHY-MAC dialogue to a BER exchange.

Should other parameters be considered in order to improve system performance?

SchedulingFairness

Traffic Modelling

Throughput

Delay

PER BER

ModulationScheme

Power Tx/Rx

TransceiverArchitecture

Diversity

Channel &Signal

Estimation

BatteryLife

Bit Rate

MAC Layer PHY Layer

Error Correcting Code

(Binomial)

Numberof Users

QoS

?

?

?

?

Parameters ExchangeParameters Exchange

Information flows between PHY and MAC levels:

BER is used for MPR computation. Active Users used for MAC efficiency. Access Set used for PHY efficiency.

MAC

PHY

BERACT.US.

ACC.SET



Further Work

PHY LevelPHY Level

CDMA System Model

Receiver Structure



Further Work

e(t)

1

21 2 M

M

d

de Ad n a a a n

d

(.)T

o

dt

(.)T

o

dt

Detection of

Active Users

Set of Active Users

^

1d

^

Md

1a

Ma

PHY LevelPHY Level

Data Demodulator Matched Filter

Active Users Detector


Design PHY level MAC level Simulation Conclusions &

Further Work

MF ||•||2 State Estimatione

( )1kn

( )kn

( )kn

Power detectionDecision based on Traffic Information

: Indicator function that takes value 1 if the kth user is active and 0 otherwise

^T T

MF Ad A Ad A n R d z

ka

Third TPL3=5

Second TPL2=7

First TPL1=4

Dynamic Queue Protocol-DQPDynamic Queue Protocol-DQP System with M users to transmit data to a

central controller. Time axis is divided into transmission periods

(TP). A TP ends when all packets generated in the

previous TP are successfully transmitted. The basic structure is a waiting queue where

all users in the system are processed in groups of Access Set size.

Based on packet user probability in one TP (qi) and the MPR matrix, the size of the access set which contains users who can access the channel in the ith TP is chosen optimally.

Transmit packets generated before 0

Transmit packets generated in (0,4]

Transmit packets generated in (4,11]



Further Work

DQP Vs MDQPDQP Vs MDQP

MDQP stands for Modified Dynamic Queue Protocol.

Central controller in DQP is capable to distinguish between: Empty slots. Successfully received packets in non-empty

slots. Central controller in MDQP is capable to

distinguish between: Empty slots. Successfully received packets in non-empty

slots. Packets lost due to collision in non-empty

slots. Nodes with empty buffers in

non-empty slots.



Further Work

45

12345

12 5

Non-empty slot. Node 3 packet

successfully received.Node 2 packet lost.

Empty slot

Non-empty slot.Nodes 2 and 4 packets successfully received.

Access Set=35

2

12345

24

Non-empty slot. Node 2 packet lost.

Node 3 packet successfully received.

Node 1 empty

Successfully received packet Packet waiting for transmission

Empty bufferPacket Lost

1

423

23 4

2

Non-empty slot.Nodes 2 and 4 packets successfully received.

Node 5 empty.

Central controller do not know whether packets from nodes 1 and 5 have collided or the buffers of these nodes were empty.

Central controller determines that nodes 1 and 5 have empty buffers.

DQP MDQP

MDQP Optimal Access SetMDQP Optimal Access Set

Ni is chosen in order to minimise the absorbing time of a finite state discrete Markov chain.

Each state (j,k) defines: j=number of unprocessed users in one slot k=number of packets sent in one slot

2,2

2,1

2,0

1,1

1,0

0,0

C2,2

C2,1

C1,1

C1,0

1

C2,0

C1,0

C1,1

1

1

Number of Users(M)=2

Access Set (N)=2

1,..,arg min [ | , ]i i i

N MN E L q N



Further Work

Access Set Vs User Packet ProbabilityAccess Set Vs User Packet Probability



Further WorkA

ccess S

et

User Packet Probability (qi)

Number of Users (M)=15SNR=10Spreading Gain (SG)=6

Packet Length (pl)=200bitsNumber of Correcting Errors(e)=2Receiver type: Matched Filter

MDQP

DQP

Throughput Vs User Packet ProbabilityThroughput Vs User Packet Probability



Further Work

SNR=10Spreading Gain (SG)=6



Th

rou

gh

pu

t

MDQP

DQP

M=15

M=10

M=5

Packet Delay Vs User Packet ProbabilityPacket Delay Vs User Packet Probability



Further Work


SNR=10Spreading Gain (SG)=6


Packet

Dela

y

MDQP

DQP

M=15

M=10

M=5

ConclusionsConclusions

Cross-Layer design concept. New idea of PHY-MAC dialogue.

Number of active users used as an additional parameter exchange between layers.

Proposal of a centralised system PHY layer with active users detector.

MAC layer with MDQP.

System improvements in terms of throughput and packet delay.


Design PHY level MAC level Simulations Conclusions

& Further Work

MPR matrix for Ad-Hoc NetworksMPR matrix for Ad-Hoc Networks

MPR must be modified considering communications in Ad-Hoc scenarios. Communications are Half-Duplex.

A node can not receive a packet while is transmitting.

A node might successfully receive a packet not intended for it.

Packet might be lost due to collision of many packets.



& Further Work

1

3

2

4

5

Packet intended for that node

Packet not intended for that node

PHY-MAC Dialogue in the IEEE802.11bPHY-MAC Dialogue in the IEEE802.11b

CSMA/CA is used. Medium is sensed by means of active user detection mechanism.

Medium is determined IDLE when Number of users sensed < Nopt for a period longer than DIFS.

After deferral, Back-off procedure adjusted depending on Number of users sensed.

Busy Medium(N. Users>=Nopt)

Back-Offprocedure

Contention Period (CP)

SIFS

PIFS

DIFS



& Further Work

PHY-MAC Dialogue in the IEEE802.11bPHY-MAC Dialogue in the IEEE802.11b

Analytical throughput expression Nopt for throughput maximisation

Modifications on the current 802.11 standard. Additional information flows:

PHY->MAC: Number of current active users. MAC->PHY: Number of users (Nopt) to consider

busy medium. Additional field in Beacon frames or broadcast

message to transmit Nopt. Carrier sense mechanism modifications

User activity detection and MUD at PHY level. Possible change in back-off procedure for

better performance. Simulations



& Further Work

Documents

PHY-MAC Dialogue with Multi-Packet Reception