An Approach Based on Reinforcement Learning for Microstrip Antenna Design
Reinforcement Learning; Q-learning; Microstrip Antennas; Return Loss; Inset-fed.
The growing demand for wireless communication for data, video and voice has become an attractive feature in the use of microstrip antennas built into portable devices, imposing more demanding antenna designs to meet the requirements of accuracy and performance. However, a careful analysis in obtaining the design parameters of antennas is essential to guarantee a proper functioning of the structure. In this context, the machine learning technique called Reinforcement Learning (RL), through the Q-learning algorithm, is applied in microstrip antennas to obtain parameters in projects. It was initially applied to the problem of impedance matching between the power line and the radiant element of a rectangular microstrip antenna, in order to determine the best value of inset-fed (y0) in the radiant element (patch). Finally, the efficacy of RL is achieved and proven through the fabrication of prototypes of the structures, followed by results measured in a specialized laboratory and compared to the simulated results.