Improved shaping of reflector antennas using a new minimax initialization strategy

Autor: Anders Eltved, Oscar Borries, Martin S. Andersen
Rok vydání: 2018
Předmět:
Zdroj: Eltved, A, Andersen, M S & Borries, O 2018, Improved Shaping of Reflector Antennas using a New Minimax Initialization Strategy . in Proceedings of 2018 International Applied Computational Electromagnetics Society Symposium . IEEE, 2018 International Applied Computational Electromagnetics Society Symposium, Denver, Colorado, United States, 25/03/2018 . https://doi.org/10.23919/ROPACES.2018.8364326
Popis: Numerical optimization is an essential part of the design process when shaping a contoured beam reflector antenna to produce a required coverage. The minimax formulation that has been used for decades generally results in a nonconvex problem, and since the evaluation of a design is costly, local optimization is typically used in practice. An inherent challenge with the minimax formulation is that it often leads to multiple local minima. As a result, the quality of the final design obtained via local optimization depends strongly on the initial design. To address the sensitivity to the initial design, we investigate the use of initialization based on a different problem formulation. In particular, we first solve a one-sided nonlinear least-squares problem to improve an initial design, and the resulting design is then used as initialization of a minimax algorithm. We use this strategy to shape a reflector antenna starting from different initial shapes, and our results show that the initialization strategy improves the coverage with five out of six initial guesses.
Databáze: OpenAIRE