Abstract:
Like other wireless systems, Orthogonal Frequency Division Multiplexing (OFDM) requires
the proper allocation of the limited resources, like total transmit power and available
frequency bandwidth, among the users to meet the users' service requirements. As a matter of
fact, adaptive resource allocation is one of the most challenging tas~s for multiuser OFDM
systems. In this dissertation, two evolutionary approaches, Genetic Algorithm (GA) and
Particle Swarm Optimization (PSO) have been applied for adaptive subcarrier and bit
allocations to minimize the overall transmit power (margin adaptation) and to maximize the
. throughput (rate adaptation) of a multiuser OFDM system. Each user will be assigned a
number of subcarriers. This allocation of subcarriers may be done through unconstrained or
I
fairly scheduled approaches. The number of bits are then calculated according to channel state
information and subcarrier arrangements. The transmit power level as well as bit rate for an
OFDM symbol are evaluated through these subcarrier and bit information. Simulation results
reveal that both the evolutionary approaches outperform the conventional static resource
allocation schemes considerably both in unconstrained and constrained cases. The results
further assert that both. the algorithms can handle large allocation I of subcarriers without
significant performance degradation. However the performance of PSO has been found to be
better than the GA in terms of execution time, simplicity and convergence.
'. The original versions of GA and PSO have been modified in different manners to provide
further improvements. All these modified versions perform relatively better than the original
versions. Furthermore the modification of PSO has been done by three different manners
where all of them perform relatively better than the original PSO as ~ell as the original and
modified versions of GA. Finally all these modified versions have been compared with the
existing algorithms. The comparison reveals the fact that the modified versions of PSO
perform relatively much better results than the previously best algorithm for higher number of
users.
