An Efficient Antenna Optimization Framework Based on the NP to P Problem

We propose a novel antenna optimization framework leveraging an NP-to-P problem approach. Initially, the complex antenna optimization mathematical model is segmented into several simpler sub-models, clarifying the relationship between design variables and performance indicators. Upon training these sub-models, a substantial dataset is rapidly generated. Subsequently, this dataset is used to train an inverse model, which contrasts with the original mathematical model in terms of input and output relations. This inverse approach enables the direct acquisition of high-quality antenna design solutions tailored to specific requirements, thereby bypassing the need for simulations. Consequently, this method transforms the intricate "NP" problem in antenna optimization into a more tractable "P" problem. These high-quality design solutions are then incorporated as foundational knowledge into the improved differential evolution algorithm (IDE) for further refinement. This integration allows IDE to bypass initial exploratory searches, effectively reducing the journey toward the optimal solution. The experimental results indicate that this method can save approximately 55% of the simulation runs compared to traditional methods, leading to more efficient antenna optimization.