1. Adaptive Resource Allocation in Multiuser OFDM Systems
Malik Meherali Saleh EE
May 05, 2005
(05/05/05)

2. Introduction Future Generation (4G) systems to provide high data rates
Issue: Inter-Symbol Interference (ISI)
Multiple Input Multiple Output (MIMO) and Orthogonal Frequency Division Multiplexing (OFDM) to the rescue…
Resource Allocation
Margin Adaptive [Wong, Cheng, Letaief and Murch, 1999]
Rate Adaptive [Jang and Lee, 2003]
Rate Adaptive + proportional fairness [Shen, Andrews and Evans, (Draft) 2004]
In my prev talk I had mentioned the use of Hybrid OFDM and MIMO systems to overcome ISI in future generation communication systems
My focus is on the way resources such as power or subcarriers are shared between multiple users of such hybrid systems.
These are 3 papers mentioned last time which talk about adaptive resource allocation within the users
the first one minimizes power while has fixed data rates for each user as a constraint
The other two max total data rate or system capacity, while have fixed total power as a constraint
I will be talking about the last two papers

3. Subcarrier Gain User 1 User 2
Channel Magnitude
Channel Magnitude
Tones
Tones
Before talking about the papers I will throw some light on the basics of OFDM
Total BW divided into N subcarrier,
Data is sent on each individual subcarriers,
The lower the gain of the subcarrier more the power needed to transmit reliable data on that subcarrier
Since total power is fixed and limited, assigning or sharing power to subcarriers is done such that maximum system capacity is achieved
For e.g. in the first case user 1 sees low gain on subcarrier where as user 2 sees a high gain. On this basis, user 2 is assigned
Subcarriers with higher gain and so the case for user 1
But in the second case user 1 does not see any high gain carriers compared to user 2
According to the allocation of jang and lee, user 1 will never be able to send any data across if the channel persists for him and u would say that is not fair
But prof evans and collegue bring in some fairness into such a system by introducing proportional fairness among users and allowing user 1 to get a fair share of the total capacity
User 1
User 2
These figures have been created for better understanding of algorithm.
They are NOT results from simulations

4. Algorithm for [Jang and Lee]
Initialization
Set Rk = 0 for k=1~K
For n = 1 to N {
Find the user k having higher channel gain
Update Rk }
For Power Allocation
Water-filling
Equal power allocation (Ptotal/N)
Rk is the kth user capacity
The algo is that for each channel n, find the user k which has the best channel gain, and then update Rk
This is subcarrier allocation
For power allocation, the paper mentions the use of water filling, which is the greedy algorithm to allocat power
Or use simple equal power over all subcarriers which furtehr reduces complexity

5. Subcarrier allocation
User 2
User 1
Channel Magnitude
Channel Magnitude
Tones
Tones
These figures have been created for better understanding of algorithm.
They are NOT results from simulations

6. Algorithm for [Prof. Evans et al]
Initialization
Set Rk = 0, Wk = {0} for k = 1 to K and A = {1,2,…,N}
Y1:Y2:…:YK
For k = 1 to K
Find n satisfying h(k,n) > h(k,j) for all j ε A
Let Wk = Wk U {n}, A = A-{n} and update Rk
The algorithm for the second method take care of users with consistent bad channel gain
Here Rk is the kth users data rate, Uk is the subcarrier allocated for user k and A is the collection of subcarriers left to be assigned
For each user 1 to K, assign it the channel for which it has maximum gain
Add that channel to Uk
Delete that channel from collection A
For each user k the algorithm assigns a subcarrier n which has maximum gain amongst the subcarriers in A
Update Rk, Wk and remove that subcarrier from A

7. Algorithm for [Prof. Evans et al] Conti..
While A != empty
Find k satisfying Rk/Yk <= Ri/Yi for all I, 1<=i<=K
For the found k, find n satisfying h(k,n) >= h(k,j) for all j ε A
For the found k and n, let Wk = Wk U {n}, A=A-{n} and update Rk
While A is not empty
Find the user k for which minimum ratio Rk to Yk
For this user k, find the highest gain subcarrier amongst the subcarrier in collection A
Update Uk, A and Rk

8. Implementation For Non Fair System For Fair System
Code Development in progress
For Fair System
‘A Low Complexity Algorithm for Proportional Resource Allocation in OFDMA Systems’ – Ian Wong et al
Linearizes power allocation
Approximates rate proportionality

9. Results
Normalized Rate Proportions
User Number (k)

10. Results Conti…

11. Future Work Implement code for first method
Compare the two methods on the following
Total Capacity
Normalized data rate for individual user
Complexity