New Micro Genetic Algorithm for multiuser detection in WCDMA

1
New Micro Genetic Algorithm for multi-user
detection in WCDMA

• AZMI BIN AHMAD
• Borhanuddin Mohd Ali, Sabira Khatun, Azmi Hassan
• Dept of Computer and Communication System
  Engineering, Faculty of Engineering
• University Putra Malaysia.

2
Schedule

• Multi-user WCDMA Tx and Rx
• GA module
• New selection method
• Result
• Discussion

3
Multiple User Data Transmission

• In a multi-user environment, signals from
  multiple user are transmitted from similar WCDMA
  transmitter (mobile station for uplink)
• These signals interfered with each other and
  resulted in Multiple Access Interference (MAI)
  and Inter-Symbol Interference (ISI).

4
Signal received

• AT the receiver, signal received consisted of
  multiple signals (multi-bit, multi-user,
  multi-path) accumulated plus interference and
  noise.

5
A single WCDMA transmitter
6
WCDMA Receiver GA module
7
At the receiver

• Multi-path signal can be solve with RAKE Receiver
  with Maximal Ratio Combining which is part of the
  receiver.
• Multi-user signal will be separated by the
  De-Scrambler and De-Spreader. The result is the
  estimation of the data which are the output of
  the Matched Filter.

8
GA Module

• The outputs of the Matched Filter is used as
  input to the GA module.
• Only the I part of the received signal will be
  used. The Q part is used as control parameters.
• The module is based on microGA type of Genetic
  Algorithms method. In this method a small
  population size is used for faster convergence.

9
new uGA selection method

• Population initializes by mutating the original
  estimated data from matched filter output.
• The individuals in the population are evaluated
  and ranked descending.
• The crossover process will used single-point
  crossing.

10
Selection (N population)

• Best-fit individual with be crossover with the
  least fit individual.
• Bestfit1 will be crossover with the Leastfit-1.
• For N population
• - 1-10, 2-9, 3-8, 4-7, 5-6

11
For Best-fit and Least-fit

• Best-fit individual is automatically selected.
• Best-fit individual is crossover with the
  least-fit individual. The better fit of the
  resulting offspring is selected.
• The least-fit individual will be discarded.

12
Rest of the populations

• From crossover of (2-9,3-8,4-7,5-6)
• Each crossover will produced 2 offspring
• 2 parent 2 offspring 4 subpopulation
• The subpopulation will be evaluated and ranked.
• The two better-fit individuals will be selected
  for new population. The other two will be
  discarded.

13
Generations

• New generation is created.
• The process will proceed for 10 generations.
• The final best-fit individual in the final
  generation will be selected as the optimized
  solution.

14
Results
15
Result..cont
16
Resultscont
17
Resultscont
18
Comparison of Computational Complexity

• Scenario K-user, one bit symbol processing
• SGA will perform in KK100
• uGA perform in K530 but as number of K
  increased the population size couldnt cover the
  whole search space
• uGA w/newSelect perform in KK/210 similar to
  uGA but better coverage of search space.

19
Conclusion

• Result didnt show much differences between uGA
  and the uGA w/newSelect
• But from error statistic the uGA w/newSelect show
  a little improvement in number of error
  corrected.
• Calculation wise the uGA w/newSelect compute in
  less time, so its suitable for use in realtime.

20
Thank You

• Questions / Suggestions