Post on 30-Dec-2015
description
Analysis of hybrid adaptive/non-adaptive multi-user OFDMA systems with imperfect channel knowledge
Alexander Kühne
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Motivation (I)
1
OFDMA Multiple access scheme for future radio systems
Offers the possibility to allocate time-frequency resources to different users
f
t
Adaptive OFDMA: Adaptive subcarrier allocation and
modulationAdvantages:Exploitation of multi-user diversityGood performance with perfect channel
knowledgeDisadvantages:– Instantaneous channel knowledge required
at the transmitter
Non-adaptive OFDMA: Fixed subcarrier allocation and
modulationAdvantages:Exploitation of frequency diversityNo instantaneous channel knowledge
at transmitter requiredDisadvantages:– No optimal channel exploitation
possible
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Motivation (I)
1
OFDMA Multiple access scheme for future radio systems
Offers the possibility to allocate time-frequency resources to different users
f
t
Adaptive OFDMA: Adaptive subcarrier allocation and
modulationAdvantages:Exploitation of multi-user diversityGood performance with perfect channel
knowledgeDisadvantages:– Instantaneous channel knowledge required
at the transmitter
Non-adaptive OFDMA: Fixed subcarrier allocation and
modulationAdvantages:Exploitation of frequency diversityNo instantaneous channel knowledge
at transmitter requiredDisadvantages:– No optimal channel exploitation
possible
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Motivation (I)
1
OFDMA Multiple access scheme for future radio systems
Offers the possibility to allocate time-frequency resources to different users
f
t
Adaptive OFDMA: Adaptive subcarrier allocation and
modulationAdvantages:Exploitation of multi-user diversityGood performance with perfect channel
knowledgeDisadvantages:– Instantaneous channel knowledge required
at the transmitter
Non-adaptive OFDMA: Fixed subcarrier allocation and
modulationAdvantages:Exploitation of frequency diversityNo instantaneous channel knowledge
at transmitter requiredDisadvantages:– No optimal channel exploitation
possible
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Motivation (II)
2
Combine both access schemes to a hybrid OFDMA system
t
f
- Resource for non-adaptive transmission
- Resource for adaptive transmission
Frequency multiplexing
Problems:
User specific imperfect channel knowledge
How to decide which user is served adaptively or non-adaptively?How to allocate the resources?How to select the applied modulation schemes?
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Outline
Assumptions
Hybrid OFDMA
3
Problem formulation
SNR threshold problem
User serving problem
Considering overhead
Performance evaluation
Conclusions
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Assumptions
Single cell scenario: One BS with nT transmit antennas and
U MSs with nR receive antennas each
TDD-OFDMA with N subcarriers
Orthogonal Space Time Block Coding (OSTBC) or Transmit Antenna Selection (TAS) at the transmitter and Maximum Ratio Combining (MRC) at the receiver
Different user demands Du
4
System assumptions
Channel modelResource unit consisting of Q subcarriers in frequency and MT OFDMA symbols in time
Temporally correlated block fading
Resource unit based Channel Quality Information (CQI):
Imperfect CQI due to time delays and estimation errors
Q
MTt
f
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Problem formulation hybrid OFDMA
5
Preprocessing:
Impairment parameters
SNR threshold vector
User serving vector
User demand vector
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Problem formulation hybrid OFDMA
6
Problem formulation:
Preprocessing:
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Problem formulation hybrid OFDMA
6
Problem formulation:
Preprocessing:
SNR threshold problem
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Problem formulation hybrid OFDMA
6
Problem formulation:
Preprocessing:
User serving problem
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Problem formulation hybrid OFDMA
6
Problem formulation:
Preprocessing:
User serving problem
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Two types of adaptive/non-adaptive resource allocation
7
Non-Adaptive First (NAF) allocationFirst, the resource units of the non-adaptive users are allocated following an round robin approach
Second, the remaining resource units are allocated following the WPFS policy
t
f
Non-adaptive user
adaptive users
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Two types of adaptive/non-adaptive resource allocation
7
Non-Adaptive First (NAF) allocationFirst, the resource units of the non-adaptive users are allocated following an round robin approach
Second, the remaining resource units are allocated following the WPFS policy
Adaptive First (AF) allocation
First, all resource units are allocated to the adaptive users applying WPFS
Second, the worst of these selected resource units are re-allocated to non-adaptive users t
f
Non-adaptive user
adaptive users
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
SNR threshold problem
8
Goal: To optimally adjust the modulation scheme SNR thresholds for each possible in dependency of the impairment parameters and the different user demands
Adjustment of the weighting factors applying WPFS to fulfill user demands
Analysis of the distribution of the SNR values of allocated resource units
Derivation of analytical expressions of the average user data rate and bit error rate (BER) using the CQI error models together with SNR distributions
Maximization of the user data rate subject to the target BER using the analytical expressions by adjusting the SNR thresholds
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
SNR threshold problem
8
Goal: To optimally adjust the modulation scheme SNR thresholds for each possible in dependency of the impairment parameters and the different user demands
Adjustment of the weighting factors applying WPFS to fulfill user demands
Analysis of the distribution of the SNR values of allocated resource units
Derivation of analytical expressions of the average user data rate and bit error rate (BER) using the CQI error models together with SNR distributions
Maximization of the user data rate subject to the target BER using the analytical expressions by adjusting the SNR thresholds
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Adjustment of the weighting applying WPFS
9
Assumptions:Each user demands Du resource units with
Users having the same demand are grouped in demand groups with i=1,..,G
Resource units for adaptive users are allocated following WPFS policy:
Question: How to adjust p such that Du is fulfilled for each user?No direct relation between pu and Du
Different antenna techniques (OSTBC-TAS) and adaptive/non-adaptive resource allocation schemes (NAF-AF)
Solution:
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Weighting for WPFS
10
Example with
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Weighting for WPFS
10
Example with
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Maximizing user data rate
For both OSTBC and TAS as well as NAF and AF analytical expressions for the average user data rate and BER of an adaptively served user u can be derived:
- SNR thresholds
- Channel estimation error variance
- Correlation coefficient (time delay)- User demand vector11
- Number of TX/RX antennas
- User serving vector
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Assumptions: Parameters are known at the BS
12
Solution:
Non-adaptive users: Problem reduces to a one-dimensional search for the proper modulation scheme
Adaptive users: Lagrange multiplier approach to determine SNR threshold vector
Maximizing user data rate by means of user-wise SNR threshold optimization
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
User serving problem
13
For each possible user serving realization , the maximum achievable data rate of adaptively and non-adaptively served users subject to the target BER are determinable
Problem: Find that which maximizes the system data rate while fulfilling the minimum user data rate requirement for NAF and AF
Assumption:
There are 2U possible user serving realizations
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Algorithms
Exhaustive Search (ES) algorithm:
14
Check all 2U possible solutions
Unpractical for a large number of users
Reduced Complexity (RedCom) algorithm:
Notice: User data rate and BER of adaptively served users do not depend on itself, but on the number of adaptive users in the different demand groups
Not all 2U combinations have to be tested but only the tuples
Example:
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Algorithms
Exhaustive Search (ES) algorithm:
14
Check all 2U possible solutions
Unpractical for a large number of users
Reduced Complexity (RedCom) algorithm:
Notice: User data rate and BER of adaptively served users do not depend on itself, but on the number of adaptive users in the different demand groups
Not all 2U combinations have to be tested but only the tuples
Example:
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Algorithms
Exhaustive Search (ES) algorithm:
14
Check all 2U possible solutions
Unpractical for a large number of users
Reduced Complexity (RedCom) algorithm:
Notice: User data rate and BER of adaptively served users do not depend on itself, but on the number of adaptive users in the different demand groups
Not all 2U combinations have to be tested but only the tuples
Example:
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Algorithms
Exhaustive Search (ES) algorithm:
14
Check all 2U possible solutions
Unpractical for a large number of users
Reduced Complexity (RedCom) algorithm:
Notice: User data rate and BER of adaptively served users do not depend on itself, but on the number of adaptive users in the different demand groups
Not all 2U combinations have to be tested but only the tuples
Example: for G=U:
for G=1:Further complexity reduction possible exploiting monotonic behavior of data rate with respect to number of adaptive users
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Complexity
15
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Considering overhead
16
So far, no pilot and signaling overhead has been taken into account
To achieve realistic results and for a fair comparison, it is important to incorporate the overhead in the overall system performance
Pilot and signaling overhead effects both downlink and uplink
Introduction of frame structure to identify the amount of required pilot and signaling overhead in hybrid systems
Introduction of an effective data rate taken into account the overhead in both up- and downlink
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Considering overhead – Superframe structure
17
LSF – superframe length
MT – time frame size in OFDMA symbols
TS – ODFMA symbol duration
for NAF LSF ≥ 1
for AF LSF = 1
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Considering overhead – Effective user data rate
18
For a given user set of one can formulate the effective user data rate of adaptively or non-adaptively served users as a weighted sum of uplink and downlink data rates:
The effective system data rate can be also maximized using the RedCom algorithm
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Simulation Parameters
19
Bandwidth for DL and UL each 10 MHz
Number U of users 15
Number N of subcarriers 240
Frequency block size Q 8
Number Nru of resource units 30
Time frame size MT in OFDMA symbols 28
Number nT of transmit antennas 2
Number nR of receive antennas each 2
Multiple antennas scheme TAS-MRC
Carrier frequency 2 GHz
Target bit error rate BERT 10-3
User demand D in resource units [2, 2, …., 2]
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Neglecting overhead (I)
20
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Neglecting overhead (I)
20
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Neglecting overhead (I)
20
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Neglecting overhead (II)
21
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Neglecting overhead (III)
22
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Considering overhead (I)
23
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Performance evaluation – Considering overhead (II)
24
18.03.10 | Alexander Kühne, Technische Universität Darmstadt, Communications Engineering Lab
Conclusions
Analytical expressions for the average user data rate and BER of a hybrid OFDMA system
25
for two different adaptive/non-adaptive resource allocation schemes NAF and AF
applying OSTBC and TAS in combination with MRC
for different user demands
assuming imperfect CQI
Maximization of system data rate subject to target BER and minimum user data rate by solving the SNR threshold and the user serving problem
Consideration of pilot and signaling overhead
Hybrid OFDMA systems outperform conventional pure adaptive or pure non-adaptive OFDMA systems for increasing user-dependent imperfect CQI even when considering overhead