| dc.description.abstract | The antenna geometry synthesis plays an important role to determine the
physical layout of the antenna array, which produces the radiation pattern closest to the actual desired pattern. The synthesis can be realized by defining the location of antenna
array elements, and by choosing suitable excitation of amplitude, and excitation phase
applied on the antenna array elements. Many synthesis techniques are done through
suppressing the side lobe level (SLL) and/or mitigating prescribed nulls while
simultaneously maintaining or improving the major lobe radiation intensity. Studies
show that some conventional analytical, numerical, and modern evolutionary algorithm
(EA) or evolutionary computation (EC) techniques have certain limitations in antenna
array geometry synthesis. This includes beamwidth expanding and directivity
saturation in amplitude tapering, exhaustive checking impairment in analytical method,
disparity predicament between local and global search accelerators in particle swarm
optimization (PSO), and drawbacks of crossover and mutation operators in genetic
algorithm (GA). This thesis presents the sequential development of enhanced and
hybrid versions of cuckoo search (CS) metaheuristic algorithm as an alternative of
EA/EC technique for symmetric linear antenna array synthesis. Firstly, the proposal of
the modified CS (MCS) algorithm through the integration with the Roulette wheel
selection operator, dynamic inertia weight, and dynamic discovery rate controlling the
best solutions exploration for a single objective (SO) optimization. Secondly, there is
the hybridization of MCS with PSO (MCSPSO), and MCS with GA (MCSGA) in both
SO and weighted−sum multiobjective (MO) approaches. Thirdly, the proposed
amalgamation of MCS with strength Pareto evolutionary algorithm (MCSSPEA), hill
climbing (HC) stochastic method within MCSSPEA algorithm (MCSHCSPEA), and
PSO within MCSSPEA algorithm (MCSPSOSPEA) equipped with distance expansion
formulae to reduce local trap problem. These newly techniques are specifically for
Pareto MO optimization to find non−dominated solutions including element location,
excitation amplitude, and excitation phase. All the tested algorithms development,
source code writing, and results execution are performed using MATLAB scientific
software. The optimal solutions are then compared against corresponding counterparts.
Based on simulation results, the proposed MCSPSO outperforms other SO and
weighted−sum MO algorithms whereas the proposed MCSPSOSPEA algorithm
surpasses other tested Pareto MO algorithms in SLL suppression and/or nulls mitigation
whilst achieving a high linear antenna directivity, and small half−power beamwidth
(HPBW), respectively. | en_US |
