A Novel Fitness Function to Enhance the Performance of Genetic Algorithm Optimization of Micro-strip Patch-antenna Dimensions for 5 GHz

Abstract

Micro-strip patch antennas (MSPAs) use standard design equations involving multiple design variables. The theoretically obtained design parameters may not conform to the desired requirements in an experimental setup. The trial-and-error method of arriving at the optimum values from the calculated values may not always be fruitful and may be difficult or even impossible when multiple variables need to be optimized. The results may not guarantee the best solution as better solutions exist. Therefore, MSPA dimensions are optimized for better results using a genetic algorithm (GA). Optimization is centered on the formulation of the fitness function. The key to success lies in adequately formulating the fitness function, considering the design variables that need to be optimized to achieve the desired output. A novel fitness function is proposed that uses graded penalties to guide a GA to converge to better results during optimization. Optimization of patch-antenna dimensions using the novel fitness function returned a patch antenna resonating at exactly 5 GHz with a bandwidth of 300 GHz while improving the return loss from - 24.16 dB to -34.79 dB. The role of correctly formed fitness functions in guiding GA towards better results was established.