multiuser resource allocation in multichannel wireless communication systems
DESCRIPTION
Multiuser Resource Allocation in Multichannel Wireless Communication Systems. Zukang Shen Ph.D. Defense Committee Members: Prof. Jeffrey G. Andrews (co-advisor) Prof. Melba M. Crawford Prof. Gustavo de Veciana Prof. Brian L. Evans (co-advisor) Prof. Robert W. Heath, Jr. - PowerPoint PPT PresentationTRANSCRIPT
Multiuser Resource Allocation in Multichannel Wireless Communication Systems
Zukang ShenPh.D. Defense
Committee Members:Prof. Jeffrey G. Andrews (co-advisor)
Prof. Melba M. CrawfordProf. Gustavo de Veciana
Prof. Brian L. Evans (co-advisor)Prof. Robert W. Heath, Jr.Prof. Edward J. Powers
Communications, Networks, and Systems AreaDept. of Electrical and Computer Engineering
The University of Texas at AustinJan. 19, 2006
(updated slides)
2
Outline
Contribution Multichannel Objective New Constraints
Low Complexity Algorithm
#1: Multiuser orthogonal
frequency division multiplexing
Frequency domain
Sum capacity
Proportional user data
rates
Decouple subchannel and power allocation
#2: Multiuser multi-antenna systems
with block diagonalization
Spatial domain
Sum capacity
Joint precoding and post-
processing
Receive antenna selection
#3: User selection in multi-antenna
systems with block diagonalization
Spatial domain
Sum capacity
Systems with a large number of
users
Greedy capacity and channel norm based algorithms
3
Resource Allocation in Wireless Systems High data rate transmission
Wireless local area networks (WLAN) 54 -- 108 Mbps Metropolitan area networks (WiMAX) ~10 -- 100 Mbps Cellular systems (3GPP) ~1 -- 4 Mbps
Limited resources shared by multiple users Transmit power Frequency bandwidth Transmission time Code resource Spatial antennas
Resource allocation impacts Power consumption User throughput System latency
user 4 user 5 user 6
user 1 user 2 user 3
time
frequency
code/spatial
4
Multiuser Diversity Multiuser wireless communication systems
Independent fading channels Multiuser diversity
0 0.1 0.2 0.3 0.4 0.5-40
-30
-20
-10
0
10
Time (sec)
Cha
nnel
gai
n (d
B)
Rayleigh Fading Channel in a 10-user System
single user gainmax user gain
Resource Allocation
Static Adaptive
Users transmission
order
Pre-determined
Smartly scheduled
Channel state
information
Not exploited
Wellexploited
SystemPerformance Poor Good
time/frequencytim
e/frequency
user 1 user 2 user 3 user K
5
Downlink Multiuser Multichannel Systems Downlink systems
Centralized basestation transmits to multiple users simultaneously
Limited resources at basestation Multiple channels created in
Frequency: orthogonal frequency division multiplexing (OFDM) Space: multiple transmit and receive antennas
Adaptive resource allocation Goal: Optimize system throughput subject to constraints Method: Formulate resource allocation as optimization problem
Optimal solution typically computationally prohibitive to findLow complexity resource scheduling algorithms desired
Assumption: Perfect channel state information of all users known at basestation
6
Outline Introduction Contribution #1: Adaptive resource allocation in multiuser
OFDM systems with proportional rate constraints Optimization framework balancing throughput and fairness Decoupling subchannel and power allocation Allocating power optimally for a given subchannel allocation
Contribution #2: Sum capacity of downlink multiuser MIMO systems with block diagonalization
Contribution #3: Low complexity user selection algorithms in multiuser MIMO systems with block diagonalization
Conclusion
7
Multiuser OFDM (MU-OFDM) Orthogonal frequency division multiplexing
Zero inter-symbol interference Parallel frequency subchannels Multiple access technology
Downlink multiuser OFDM Users share subchannels and basestation transmit power Users only decode their own data
Resource allocation methods Static: TDMA, FDMA Dynamic: multiuser diversity
Users feedback channelinformation to basestation
Basestation determinesresource allocation
frequency
gain
8
MU-OFDM Adaptive Resource Allocation
Objective Advantage Disadvantage
Max sum capacity
[Jang et al., 2003]
Best sum capacity
No data rate fairness among users
Max minimum user’s capacity [Rhee et al., 2000]
Equal user data rates
Inflexible user datarates distribution
Max weightedsum capacity
[Cendrillon et al., 2004]
Data rate fairness
adjustable by varying
weights
No guarantee for required proportional
user data rates
: user k’s capacity (bits/s/Hz) as continuous function for single cell
9
MU-OFDM with Proportional Rates Objective: Sum capacity
Constraints Total transmit power No subchannel shared by multiple users Proportional rate constraints
Advantages In theory, fill gap of max sum capacity & max-min capacity In practice, allow different service privileges and different pricing
B Transmission bandwidth
K # of users
N # of subchannels
pk,npower in user k’s
subchannel n
hk,nchannel gain of user k’s
subchannel n
N0 AWGN power density
Rk User k’s capacity
System parameter for proportional rates
Contribution #1
10
Subchannel Allocation Modified method of [Rhee et al., 2000], but we keep the
assumption of equal power distribution on subchannels1. Initialization (Enforce zero initial conditions)
Set , for . Let
2. For to (Allocate best subchannel for each user)a) Find satisfying for allb) Let , and update
3. While (Then iteratively give lowest rate user first choice) a) Find satisfying for allb) For the found , find satisfying for allc) For the found and , Let , and
update
Contribution #1
11
Power Allocation for a Single User Optimal power distribution for user
Order Water-filling algorithm
How to find for
Contribution #1
K # of users
N # of subchannels
pk,npower in user k’s nth assigned subchannel
Hk,nChannel-to-noise ratio in
user k’s nth assigned subchannel
Nk# of subchannels
allocated to user k
Pk,totTotal power allocated to
user k
subchannels
Water-level
12
Power Allocation among Many Users Use proportional rate and total power constraints
Solve nonlinear system of K equations: /iteration Two special cases
Linear case: , closed-form solution High channel-to-noise ratio: and
Contribution #1
where
13
Comparison with Optimal SolutionContribution #1
-10 -5 0 5 100
0.5
1
1.5
2
2.5
3
3.5
10*log10(1/2)
Ove
rall
capa
city
(bits
/s/H
z)
optimal, E(ch1)/E(ch2)=1decoupled, E(ch1)/E(ch2)=1optimal, E(ch1)/E(ch2)=0.1decoupled, E(ch1)/E(ch2)=0.1optimal, E(ch1)/E(ch2)=10decoupled, E(ch1)/E(ch2)=10
14
Comparison with Max-Min Capacity Contribution #1
8 10 12 14 160.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Number of users K
Min
imum
Use
r's C
apac
ity (b
it/s/
Hz) proposed
max-min:equal powerTDMA
15
Comparison with Max Sum CapacityContribution #1
0 1 2 3 4 5 6 76
7
8
9
10
Fairness Index m
Erg
odic
Sum
Cap
acity
(bits
/s/H
z)
max sum capacitysingle user (higher SNR)proposedstatic TDMAsingle user (lower SNR)
1 2 3 4 5 6 7 80
0.2
0.4
0.6
0.8
Nor
mal
ized
Erg
odic
Cap
acity
Per
Use
r
User Index k
ideal, m=3proposed m=3max sum capacitystatic TDMA
16
Summary of Contribution #1 Adaptive resource allocation in multiuser OFDM systems
Maximize sum capacity Enforce proportional user data rates
Low complexity near-optimal resource allocation algorithm Subchannel allocation assuming equal power on all subchannels Optimal power distribution for a single user Optimal power distribution among many users with proportionality
Advantages Evaluate tradeoff between sum capacity and user data rate
fairness Fill the gap of max sum capacity and max-min capacity Achieve flexible data rate distribution among users Allow different service privileges and pricing
17
Outline Introduction Contribution #1: Adaptive resource allocation in multiuser
OFDM systems with proportional rate constraints Contribution #2: Sum capacity of downlink multiuser MIMO
systems with block diagonalization Block diagonalization with receive antenna selection Sum capacity of BD vs. DPC for given channels Upper bound on the ratio of DPC and BD sum capacity in Rayleigh
fading channels Contribution #3: Low complexity user selection algorithms
in multiuser MIMO systems with block diagonalization Conclusion
18
Multi-Antenna Systems Exploit spatial dimension with multiple antennas Improve transmission reliability – diversity
Combat channel fading [Jakes, 1974] Combat co-channel interference [Winters, 1984]
Increase spectral efficiency – multiplexing Multiple parallel spatial channels created with multiple antennas at
transmitter and receiver [Winters, 1987] [Foschini et al., 1998] Theoretical results on point-to-point multi-input multi-output (MIMO)
channel capacity [Telatar, 1999]
Tradeoff between diversity and multiplexing Theoretical treatment [Zheng et al., 2003] Switching between diversity and multiplexing [Heath et al., 2005]
19
MIMO Gaussian Broadcast Channels Duality with multiple access channels [Vishwanath et al., 2003]
Dirty paper coding (DPC) [Costa, 1983] Sum capacity achieved with DPC [Vishwanath et al., 2003] Iterative water-filling algorithm [Yu et al., 2004] [Jindal et al., 2005]
Capacity region [Weingarten et al., 2004]
Coding schemes approaching DPC sum capacity[Zamir et al., 2002] [Airy et al., 2004] [Stojnic et al., 2004] Too complicated for cost-effective implementations
+
+
+
+
BroadcastChannel
MultipleAccessChannel
20
Block Diagonalization (BD) Linear precoding technique
Zero inter-user interference [Spencer et al., 2004]
in the null space of Advantages: Simple transceiver design
Effective point-to-point MIMO channel Disadvantages: Suboptimal for sum capacity
Channel energy wasted for orthogonalizing user channels Transmit signal covariance matrices not optimal
21
BD with Receive Antenna Selection Why joint processing?
Confine to be selection matrix, e.g. Lower system overhead for conveying BD with receive antenna selection Exhaustive search for optimal selection matrices
Contribution #2
22
BD vs. DPC: Given Channels
Theorem: The ratio of DPC sum capacity over BD is bounded by Ratio of DPC sum capacity over TDMA bounded by
[Jindal et al., 2005] TDMA only serves one user at a time BD supports multiple users: Valid for any SNR, , , and
Lemma: If user channels are orthogonal, then
Lemma: If and user channelsare in same vector space
Contribution #2
23
BD vs. DPC: Rayleigh Fading Channels Lower bound on BD ergodic sum capacity
Fix a subset of users to serve Each user’s effective channel still Rayleigh Equal power allocated for every MIMO eigenmode
Upper bound on DPC ergodic sum capacity Allow user cooperation (effectively point-to-point channel) Cooperative channel Space-time water-filling for effective cooperative MIMO channel
Upper bound on ratio of DPC and BD ergodic sum capacity Easy evaluation with numerical integrations Bound is tight for
Medium to high SNR, or
Contribution #2
24
Simulation ResultsContribution #2
-20 0 20 40 601
1.5
2
2.5
3
SNR (dB)Erg
odic
Sum
Cap
acity
Gai
n: D
PC
vs.
BD
Proposed boundMonte Carlo, DPC/BD w RxASMonte Carlo, DPC/BD w/o RxAS
-20 0 20 40 600
50
100
150
200
SNR (dB)
Erg
odic
Sum
Cap
acity
(bits
/s/H
z) DPCBD w RxASBD w/o RxAS
25
Simulation ResultsContribution #2
6 8 10 12 14 16 18 200
10
20
30
40
50
# of Transmit Antennas Nt
Erg
odic
Sum
Cap
acity
(bit/
s/H
z) DPCBD w RxASBD w/o RxAS
SNR 20 dB
SNR 10 dB
SNR 0 dB
6 8 10 12 14 16 18 201
1.1
1.2
1.3
1.4
# of Transmit Antennas NtE
rgod
ic S
um C
apac
ity G
ain:
DP
C v
s. B
D
Proposed boundMonte Carlo, DPC/BD w RxASMonte Carlo, DPC/BD w/o RxAS
SNR=20 dB
26
Summary of Contribution #2 Sum capacity in downlink multiuser MIMO systems with
block diagonalization Formulated joint transmitter precoding and receiver post-processing
(shown in dissertation) Combined block diagonalization with receive antenna selection
Block diagonalization vs. dirty paper coding Sum capacity for given channels Ergodic sum capacity in Rayleigh fading channel
Block diagonalization achieves a significant part of the optimal sum capacity
27
Outline Introduction Contribution #1: Adaptive resource allocation in multiuser
OFDM systems with proportional rate constraints Contribution #2: Sum capacity of downlink multiuser MIMO
systems with block diagonalization Contribution #3: Low complexity user selection algorithms
in multiuser MIMO systems with block diagonalization Capacity based user selection Channel Frobenius norm based user selection
Conclusion
28
Need of User Selection for BD Zero inter-user interference requires in null space of
Dimension of : Maximum number of simultaneous users:
Assuming active users utilize all receive antennas Select subset of users to maximize total throughput Exhaustive search
Optimal for total throughput Computationally prohibitive
Related work Semi-orthogonal user set construction [Yoo et al., 2005] Antenna selection [Gharavi-Alkhansari et al., 2004]
29
Greedy User Selection AlgorithmsContribution #3
Capacity based algorithm(c-algorithm)
Channel norm algorithm(n-algorithm)
, apply BD to
users selected
Yes
No users selected or sum capacity decreases
apply c-algorithmto select subset
No
Yes
30
Computational Complexity
Proposed algorithms have complexity Average CPU run time
(Pentium M 1.6G Hz PC)
(m x n) complex matrix operation Flop counts
Frobenius norm
Gram-Schmidt orthogonalization
Water-filling algorithm
Singular value decomposition
Contribution #3
3 10 20 30 40 500
20
40
60
80
100
120
140
# of Total Users K
Tim
e (m
illis
econ
ds)
BD capacity algorithmBD channel norm algorithm
31
Monte Carlo ResultsContribution #3
2 10 20 30 40 500
5
10
15
20
25
30
35
# of Total Users K
Erg
odic
Sum
Cap
acity
(bits
/s/H
z) DPCBD OptimalBD c-algorithmBD n-algorithmBD no selection
SNR 20 dB
SNR 10 dB
SNR 0 dB
32
Summary of Contributions
Adaptive resource allocation in multiuser OFDM Balanced throughput and proportional user data rates Derived optimal power allocation given subchannel allocation
Sum capacity of downlink multiuser MIMO systems Combined block diagonalization with receive antenna selection Analyzed sum capacity of BD vs. DPC for given channels Derived upper bound on ratio of DPC and BD sum capacity in
Rayleigh fading channels
Low complexity user selection algorithms in multiuser MIMO systems with block diagonalization Proposed two algorithms with linear complexity in no. of total users Achieved near-optimal sum capacity