基于贝叶斯算法的天线端到端优化

在确定天线的拓扑结构以后, 通常需要对天线的结构参数开展反复的优化才能达到设计目标, 快速有效的优化算法有利于缩短天线的设计周期。在建立综合目标函数的基础上, 同时考虑天线的多个优化目标和限制条件, 使用贝叶斯优化算法对天线进行端到端优化。基于在线更新的数据集, 高斯过程估计出目标函数的后验分布, 进而使用获得函数进行迭代。通过两种天线模型对提出的优化算法进行仿真验证, 结果表明, 由于建立了天线参数到综合目标函数的映射关系, 整个优化过程以端到端的方式实现, 与传统的优化方法相比, 所提算法的优化结果和优化速度都具有明显的优势。

End-to-end antenna optimization based on Bayesian optimization algorithm

Once the topology of an antenna is determined, optimization of various antenna parameters is conducted to achieve the desired objectives. A fast and effective optimization algorithm is significant to shorten the antenna design period. With the use of a comprehensive objective function which fuses multiple targets and constraints into one single function, the antenna is optimized end-to-end by employing Bayesian optimization algorithm. Based on the online updated dataset, the Gaussian process estimates the posterior distribution of the comprehensive objective function, and afterwards acquisition function is utilized for the iteration. To demonstrate the proposed method, two antenna models are simulated and optimized. The results show that because the proposed method is implemented in an end-to-end way through a mapping from the antenna parameters to a comprehensive objective function, it can achieve better optimization results and higher optimization efficiency compared with the conventional methods.