Accelerated Antenna Optimization Using Gradient Search with Selective Broyden Updates

The paper addresses design optimization of antennas using accelerated gradient-based search. In our approach, finite differentiation conventionally used to estimate the antenna response Jacobian is replaced, for selected variables, by rank-one Broyden updates. Variable selection depends on the alignment between direction of the recent design relocation and the coordinate system axes. Comprehensive numerical validation indicates significant computational savings that can be achieved without compromising the design quality.