Opportunistic Scheduling in Wireless Networks

1
Opportunistic Scheduling in Wireless Networks

• Mohammed Eltayeb
• Obaid Khattak

2
Project Outline

• This report gives an overview of different
  scheduling algorithms, from the simple round
  robin algorithm, to opportunistic scheduling
  algorithms considering QoS, with simulation of
  system
• capacity
• feedback load and
• fairness.
• We divided the algorithms into fair, semi-fair
  and greedy algorithms.
• All simulations are done with Matlab 7.0 with an
  average SNR of 15dB and 1000 Ts for 30 users.

3
Back Ground Theory

• A scheduling system is implemented both in the
  mobile station (MS) and in the base station (BS).
• The BS uses a TDMA scheme and during one time
  slot, only one user can receive or transmit, and
  this user is selected by the scheduler.

4
Fair Algorithms

• Round Robin
• The RR scheduler is the simplest scheduling
  algorithm, and it is not opportunistic.
• When a user connects to the base station (BS),
  it is given a position in the queue of users, and
  the scheduler will iterate through the queue.

5
Fair Algorithms - RR
6
Fair Algorithms - RR
7
Fair Algorithms

• Opportunistic Round Robin (ORR)
• The ORR algorithm is a Round Robin scheduler.
• Channel conditions are taken into account.
• The scheduler iterates the list of users, and
  every time the best user is selected and removed
  from the list.

8
Fair Algorithm - ORR
9
Fair Algorithm - ORR
10
SEMI-FAIR SCHEDULING ALGORITHMS

• EXAMPLES AND PERFORMANCE

11
Semi-Fairness

• Middle ground between Fair Greedy
• Provide Fairness in terms of scheduling outage
• Feedback load not zero but not rate optimal
  either
• Example Switched Diversity Scheduling (SDS)

12
SDS

• Family of algorithms based on multi-antenna
  systems schemes
• Specific Threshold ?th is set
• Scans users to find CNR gt ?th
• If user found, selected
• At each time slot, sequence may be randomized or
  organized in special way
• Examples
• Selection Combining Transmission (SCT)
• SET with Post-Selection (SETps)

13
SCT

• Checks ALL users, selects user with highest CNR
• Fair if all users are i.i.d
• Advantage
• Only form of SDS which is rate optimal
• Disadvantage
• Normalized feedback load (NFL) unity

14
MASSE Performance of SCT
15
Throughput Fairness in SCT
16
SETps

• Extension of Switch-and-Examine Transmission
  (SET)
• First scanned user with CNR gt ?th selected
• If no user CNR gt ?th ? User with greatest CNR
  selected
• Combats scheduling outage
• At each time slot, list randomized
• Provides level of fairness

17
MASSE of SETps
18
Throughput Fairness of SETps
19
Time-slot Fairness of SETps
20
NFL of SETps
21
GREEDY SCHEDULING ALGORITHMS

• EXAMPLES AND PERFORMANCE

22
Greedy Algorithms

• More concerned with maximizing system throughput,
  not fairness to individual users
• Do provide fairness when all users have i.i.d.
  channel conditions
• Rate optimal, MASSE values equal
• Examples
• Maximum CNR Scheduling (MCS)
• Optimal Rate, Reduced Feedback (ORRF)

23
MCS

• All users report their CNR to BS
• User with best channel selected
• Rate optimal
• Large overhead in reporting CNR values
• Normalized feedback load (NFL) unity
• Poor throughput and time-slot fairness
• Same as SCT

24
MASSE of optimal schedulers
25
Optimal Rate, Reduced Feedback (ORRF)

• Scheduler decides threshold CNR
• Distributed to all users
• Users with CNR gt Threshold reply
• Best user selected
• If no user replies
• Scheduler requests full feedback
• Every user returns CSI (Channel State
  Information)
• After full feedback or without it, best user
  selected

26
NFL of ORRF
27
Time-slot Fairness of ORRF
28
Throughput Fairness
29
MASSE-based Comparison
30
NFL-based Comparison
31
References

• 1 P. Viswanath, D. N. C. Tse, and R. Laroia,
  _Opportunistic beamforming
• using dumb antennas,_ IEEE Trans. Inform.
  Theory, vol. 48, pp. 1277_
• 1294, June 2002.
• 2 A. J. Goldsmith and P. P. Varaiya, _Capacity
  of fading channels with channel
• side information,_ IEEE Trans. Inform. Theory,
  vol. IT-43, pp. 1896_
• 1992, Nov. 1997.
• 3 D. Gesbert and M.-S. Alouini, _How much
  feedback is multi-user diversity
• really worth?,_ in IEEE Int. Conf. on
  Communications (ICC'04), (Paris,
• France), pp. 234_238, June 2004.
• 4 V. Hassel, M.-S. Alouini, G. E. Øien, and D.
  Gesbert, _Rate-optimal multiuser
• scheduling with reduced feedback load and
  analysis of delay effects._
• Submitted to IEEE Int. Conf. on Comm. (ICC'05),
  (Seoul, South Korea),
• May 2005.
• 5 M. Johansson, _Issues in multiuser
  diversity._
• http//www.signal.uu.se/Research/PCCWIP/Visbyrefs
  /Johansson_Visby04.pdf.
• Presentation at WIP/BEATS/CUBAN workshop Wisby,
  Sweden, Aug.
• 2004.
• 6 R. Knopp and P. A. Humblet, _Information
  capacity and power control in
• single cell multiuser communications,_ in IEEE
  Int. Conf. on Communications