Date post: | 18-Jan-2016 |
Category: |
Documents |
Upload: | carrington |
View: | 42 times |
Download: | 0 times |
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Smart Contention Resolution Random Access Wireless Networks With Unknown Multiple
Users: A Joint Layer Design Approach
Xiaohua Li and Wenyu Liu Department of Electrical and Computer Engineering, State University of
New York at Binghamton, Binghamton, NY 31902Motorola Inc., Libertyville, IL 60048
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Outline
IntroductionJoint PHY/MAC Layer design
Performance AnalysisSimulationsConclusions
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Introduction
Issues:• Wireless spectrum is becoming scarce• Advanced access schemes required to improve efficiency to improve throughput to support packet priority/QoS • Random channel access is more preferred • Need resolve contentions
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Introduction (cont.)
Issues: (cont.)• Contention resolution is challenging in wireless networks• Traditional ways: operate in MAC layer, collided signals discarded instead of utilized• Signal processing ways: separate collided signal in PHY layer
Questions:Benefit from both PHY/MAC layer?
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Introduction (cont.)
Proposed Solution:• Joint PHY/MAC layer design
– PHY: signal separation or multiuser detection– MAC: smart packet scheduling
• Compared with traditional ways: – collided signal is used instead of discarded
• Compared with other signal processing ways– Preserves packet priority, support QoS – Works with unknown multiple users– Computationally efficient and robust, i.e., does not
suffer ill-channel or near-far conditions.
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
System Model of Wireless Networks• The wireless network has a central controller and multiple unknown mobile users:• Central controller: PHY collision separation, MAC packet scheduling
User n
User 1 User 2
Controller(User 0)
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Packet Flow in the Wireless Networks
… Access request packets
(Controller & Users)
Data packets(Users)
Access request packets
(Controller & Users)
Data packets(Users) …
Problem: if there are too many unknown users, contentions in access request slots are high
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Packet Flow in the Wireless Networks (cont.)
• Solutions: divide and conquer
• Procedure:– Each active user generate I random number– Each random number is corresponding to a unique
access request packet– Actives users are grouped according to I or access
request packets
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Random Access Scheduling Protocol 1. Central controller ask all active users for access
requests2. Active users generate random numbers
• Form subgroups distributedly by the numbers• Transmit access request packets according to the random
numbers
3. Central controller detects collided packets• If there are active users from different subgroups, there
access request packets collide• Collided packets can be detected successfully• From collided signal, central controller know active
subgroups• Central controller asks each subgroup one-by-one to
request again
4. Go to 2
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Random Access Scheduling Protocol (cont.)• Packet priority:
– Users generate random numbers according to priority– Central controller gets priority information after separating
collided access requests – Central controller arrange active users to transmit data packets
according to their priorities.
• Efficiency:– Contention exists in access request slots only– If active users have different random integer, no collision
happens for data packet transmission– Size of access request slots is reduced by divide-and-conquer
• Collision detection:– Orthogonal codes can be used because of the reduced access
request slot size
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Access Request Collision Resolution
• Key point: need K different access request packets (codes) only
– One codes for each group instead of user– All active users in a group transmit the same code, and will
be divided in the next turn
• Each user generate random number – find from
– Select the corresponding code to transmit as access request
packet
ja
ija ,
1
0,
I
i
iijj Kaa
1,,0)],(,),0([ KkMcc kkTk c
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Access Request Collision Resolution
• There are at most K different collided access codes, although there are unknown number of users
• Received collided signal
kk
Hk
kk
kk
k jjjjj
T
NMcc
McNc
Lhh
MxNxN
Ch
00
x
)()0(
)()(
]),()0(,[
)]()([)(
21
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Access Request Collision Resolution
issionfar transm-near
us,asynchrono multipath, in even codes, orthogonalby
ly /efficientcompletely cancelled ceinterferen access-Multi
active is group if,
activenot is group if,)(
metric Decision
if,
if, that such detectorsK find can we
,, matrices symboldifferent most at With 10
k
kNz
kl
kl
K
Hk
kH
k
kl
K
1h
0fx
0
1fC
CC
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Access Request Collision Resolution
• Active group/user detection (collision separation)– Construct received sample vector – Apply detector to calculate decision
metrics – Compare with threshold to determine
whether each group k is active (feedback to MAC)
kf
kz
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Performance Analysis
Probability of Data Packet Collision:
where A is the maximum range of the random number, u is the number of active users contending for channel access at the same time
A
iAuP
u
ic
1
01)(
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Performance Analysis (cont.)
System Efficiency:
ratio of successful to total transmission in one frame instantaneous throughput
Without detection error:
With detection error
acd
cdc
ad
d
ILuPuL
uPuLuEuP
ILL
L
)(1(
))(1()]([))(1(
ad
d
ad
d
ILuL
uLu
ILL
L
)(
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Performance Analysis (cont.)
aad
d
LL
L
0
System Throughput:
Packet delay:
],)1
([
: withinis onelast theofdelay the
, :ispacket first theofdelay the
,contendingpacket data if
d
1
1d
d
aa
I
ii
a
IuuK
uIu
I
u
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Simulations
DER versus SNR, for offered load 0.5 and 1:satisfactorily low detection errors for even low SNR
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Simulations (cont.)
Throughput vs. offered traffic load :Achieve much higher throughput than ALOHA and CSMA.
Especially, if the DER is not too low, it achieves almost the offered traffic load up to unity.
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Simulations (cont.)
Packet delay vs. throughput:Has better (much smaller) packet delay than the ALOHA. With DER=0, 0.01,
0.05.
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Conclusions
– Use signal processing, specifically, multiuser detection principles, to resolve random access contentions.
– Central controller schedules random access request packet detection
– Collisions are resolved through signal separation in systems with unknown multiple users
– Enhance throughput, support QoS
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
-Questions? Questions?
-Issues?Issues?
-Comments?Comments?
ICC 2003 - May 11, 2003 A Joint Layer Design X. Li, W. Liu
Thank You !!!
Xiaohua LiDepartment of Electrical and Computer Engineering
State University of New York at Binghamton
Binghamton, NY 12309
Email: [email protected]
Wenyu LiuPersonal Communication Sector
Motorola Inc.
Libertyville, IL 60048
Email: [email protected]