Transmission Rate Scheduling with Fairness Constraints in Downlink of CDMA Data Networks

1
Transmission Rate Scheduling with Fairness
Constraints in Downlink of CDMA Data Networks

• AuthorEuntaek Lim and Sehun Kim
• SourceIEEE Transactions on Vehicular, Vol.54, 
  No.1, Jan. 2005
• SCIRank 22/57 38.6 (Impack Factor 0.611)
• Reporter Tsang-Yuan Tsai (???)
• Date2005/9/3

2
Outline

• Introduction
• One-by-one transmission-rate scheduling methods
• TSMABE (Time-Span Minimization and Best Effort)
• TRSSFC (Transmission-Rate Scheduling with Soft
  Fairness Constraints)
• Numerical Simulations
• Conclusion
• Comment

3
motivation

• The total throughput of the downlink in a CDMA
  system can be maximized by choosing the user with
  the best signal-to-interference ratio (SIR) at
  each time instance, but the scheme is with no
  fairness.
• How to proposed the scheme for the throughput
  maximization problem with fairness among users in
  the downlink of a CDMA system

4
Celullar mobile

• The blue cell is the intracell for the ms1, the
  other color cells are the intercells
• for the ms1
• The green cells are the neighboring
• cells for the blue cell

5
Introduction(1/3)

• Signal-to-interference ratio (SIR) for user i
• SIRi
• where PiGi is the received power, Iiinter is
  the interference from the intercell, Iiintra is
  the interference from the intracell and ?is the
  noise
• The maximum transmission rate ri

Where W is chip rate and is the
bit-energy-to-noise-density ratio that provides
the minimum acceptable bit-error rate (BER)
6
Introduction(2/3)

• One-by-one transmission-rate scheme means only
  one user use the maximun power transmitted the
  data in the whole cell, and there is no intracell
  interference so the transmission rate of the user
  s
• The throughput R of the cell during short time
  interval ?t is

7
Introduction(3/3)

• The base station transmits to user i or not at
  time slot t is indicated by binary decision
  variable xit (0 or 1). Rewrite the throughput for
  user i during time slot t
• The total throughput of user i during the whole
  time duration T (1N time slots)

8
Transmission-Rate Scheduling with hard Fairness
constraints(1/2)
Where Ai is the minimun throughput requirement
for user i during The whole time duration T and
E is the set of active users in the cell.
9
Transmission-Rate Scheduling with hard Fairness
constraints(2/2)

• Let
  gt Rit Ditxit, so rewrite the
  constraints

10
Time-Span Minimization and Best EffortTSMABE

• Let F be the set of the users who need more time
  slots to satisfy minimun throughput requirements
  of all users.

11
TSMABE-Maximun throughputfor remaining time slots
12
Example of TSMABE(1/4)

• Provided three users and the location is fixed
  and SIR is fixed for each user. The minimun
  throughput requirement A 800bits for each user.
  And the whole time duration is 7 time slots.
• Time slot 1 F 1,2,3
• A1 800 500 300

13
Example of TSMABE(2/4)

• Time slot 2 F 1,2,3
• A1 300 500 -200 0 gt F F 1
  2,3
• Time slot 3 F 2,3
• A2 800 450 350 ? 0

14
Example of TSMABE(3/4)

• Time slot 4 F 2,3
• A2 350 450 -100 0 gt F F 2
  3
• Time slot 5 F 3
• A3 800 400 400

15
Example of TSMABE(4/4)

• Time slot 6 F 3
• A3 400 400 0 0 gt F F 3
• Time slot 7 Best effort method( the highest SIR
  value is selected)

16
Opportunity Cost Method and Best EffortOCMABE
17
OCMABE-Maximun throughputfor remaining time slots
18
Example of OCMABE(1/4)

• Provided all time slots estimated as the
  following
• and the minimum requirements
  A1A2A3800bits, F1,2,3 , T 1,2,3,4
• Iteration 1

T T - 2 1,3,4 , A3 800 900 -100
0 F F 3 1,2
19
Example of OCMABE(2/4)

• Iteration 2 F 1,2 , T 1,3,4

T T - 1 3,4 , A1 800 700 100
20
Example of OCMABE(3/4)

• Iteration 3 F 1,2 , T 3,4

T T - 3 4 , A1 100 600 -500 0 F
F 1 3
21
Example of OCMABE(4/4)

• Iteration 4 F 2 , T 4

T T - 4 , A2 800 500 300 F F
1 3 Not all user match the minimum
requirement , but T will result in the
solution infeasible.
22
Transmission-Rate Scheduling with Soft Fairness
ConstraintsTRSSFC
23
Example of TRSSFC(1/4)

• Provided three users and the location is fixed
  and SIR is fixed for each user. The minimum
  throughput requirement A 800bits for each user.
  And the whole time duration is 7 time slots. ß
  1.5 .
• Time slot 1 F 1,2,3

If (A1 800) gt 0 then A1 800 500 300
24
Example of TRSSFC(2/3)

• Time slot 2 F 1,2,3
• Time slot 3 F 1,2,3

Because the user 1 lt 1.5 ,so user 2 is
selected If (A2 800) gt 0 then A2 800 450
350
If (A1 300) gt 0 then A1 300 500 -200 If
(-200) 0 then F F 1 2,3
25
Example of TRSSFC(2/3)

• Time slot 4 F 2,3
• Time slot 5 F 3

If (A2 350) gt 0 then A2 350 450 -100 If
(-100) 0 then F F 2 3
If (A3 800) gt 0 then A3 800 400 400
26
Example of TRSSFC(2/3)

• Time slot 6 F 3
• Time slot 7 F

If (A3 400) gt 0 then A3 400 400 0 If (0)
0 then F F 3
After the time slot the highest is always
selected. It is the same as the best effort
method.
27
Numerical Simulations

• Assume
• The location of each user is fixed
• The position of each user is randomly generated
  assuming that users are probabilistically
  uniformly distributed in the cell
• Each time slot the neighboring cell transmits
  with random power
• The path loss is 4
• The cell radius is 1 km
• The noise level is -150dBW
• W 1.229MHz
• Let Rm be the throughput for all the duration by
  best effort method. aRm will be the minimum
  requirement and where a is the value between 0
  and 1.

28
PERFORMANCE OF TSMABE, TRSSFC, AND OCMABE WHEN
a 0.4 AND ß 1.5
29
PERFORMANCE OF TSMABE, TRSSFC, AND OCMABE WHEN
a 0.5 AND ß 1.5
30
Outage probabilities of three methods for
different values of a

• The more time slot the lower probability

31
Throughput for different values of ß
32
Conclusion

• The intercell interference is assumed to be
  random.
• Two proposed models
• A model with hard minimum requirement constraints
  TSMABE
• A model with soft minimum requirement constraints
  - TRSSFC
• The proposed scheduling methods require the
  information about only the current SIR

33
Comment

• TRSSFC is more fairness than TSMABE. But the
  performance metric is only the throughput. So the
  model balancing the fairness and the throughput
  maybe need in study.

34
A users estimator for the future throughput

• is an estimator of the throughput of user at
  a future time slot.
• If gt then we assign this time slot
  to user k, since this slot has a high throughput
  for user k compared to the users expected future
  throughputs and vice versa.
• Let ß be a parameter that is greater than 1.